@pydantic/genai-prices 0.0.61 → 0.0.62

package/dist/cli.js

@@ -172,7 +172,7 @@ class De {
     let L = null;
     Ae();
     let V = [];
-    const w = Object.assign(/* @__PURE__ */ Object.create(null), { _: [] }), se = {};
+    const b = Object.assign(/* @__PURE__ */ Object.create(null), { _: [] }), se = {};
     for (let e = 0; e < r.length; e++) {
       const a = r[e], c = a.replace(/^-{3,}/, "---");
       let n, t, l, p, d, v;
@@ -182,39 +182,39 @@ class De {
         Y(a);
         continue;
       } else if (a.match(/^--.+=/) || !k["short-option-groups"] && a.match(/^-.+=/))
-        p = a.match(/^--?([^=]+)=([\s\S]*)$/), p !== null && Array.isArray(p) && p.length >= 3 && (g(p[1], m.arrays) ? e = K(e, p[1], r, p[2]) : g(p[1], m.nargs) !== !1 ? e = j(e, p[1], r, p[2]) : b(p[1], p[2], !0));
+        p = a.match(/^--?([^=]+)=([\s\S]*)$/), p !== null && Array.isArray(p) && p.length >= 3 && (g(p[1], m.arrays) ? e = K(e, p[1], r, p[2]) : g(p[1], m.nargs) !== !1 ? e = j(e, p[1], r, p[2]) : w(p[1], p[2], !0));
       else if (a.match(U) && k["boolean-negation"])
-        p = a.match(U), p !== null && Array.isArray(p) && p.length >= 2 && (t = p[1], b(t, g(t, m.arrays) ? [!1] : !1));
+        p = a.match(U), p !== null && Array.isArray(p) && p.length >= 2 && (t = p[1], w(t, g(t, m.arrays) ? [!1] : !1));
       else if (a.match(/^--.+/) || !k["short-option-groups"] && a.match(/^-[^-]+/))
-        p = a.match(/^--?(.+)/), p !== null && Array.isArray(p) && p.length >= 2 && (t = p[1], g(t, m.arrays) ? e = K(e, t, r) : g(t, m.nargs) !== !1 ? e = j(e, t, r) : (d = r[e + 1], d !== void 0 && (!d.match(/^-/) || d.match(I)) && !g(t, m.bools) && !g(t, m.counts) || /^(true|false)$/.test(d) ? (b(t, d), e++) : b(t, R(t))));
+        p = a.match(/^--?(.+)/), p !== null && Array.isArray(p) && p.length >= 2 && (t = p[1], g(t, m.arrays) ? e = K(e, t, r) : g(t, m.nargs) !== !1 ? e = j(e, t, r) : (d = r[e + 1], d !== void 0 && (!d.match(/^-/) || d.match(I)) && !g(t, m.bools) && !g(t, m.counts) || /^(true|false)$/.test(d) ? (w(t, d), e++) : w(t, R(t))));
       else if (a.match(/^-.\..+=/))
-        p = a.match(/^-([^=]+)=([\s\S]*)$/), p !== null && Array.isArray(p) && p.length >= 3 && b(p[1], p[2]);
+        p = a.match(/^-([^=]+)=([\s\S]*)$/), p !== null && Array.isArray(p) && p.length >= 3 && w(p[1], p[2]);
       else if (a.match(/^-.\..+/) && !a.match(I))
-        d = r[e + 1], p = a.match(/^-(.\..+)/), p !== null && Array.isArray(p) && p.length >= 2 && (t = p[1], d !== void 0 && !d.match(/^-/) && !g(t, m.bools) && !g(t, m.counts) ? (b(t, d), e++) : b(t, R(t)));
+        d = r[e + 1], p = a.match(/^-(.\..+)/), p !== null && Array.isArray(p) && p.length >= 2 && (t = p[1], d !== void 0 && !d.match(/^-/) && !g(t, m.bools) && !g(t, m.counts) ? (w(t, d), e++) : w(t, R(t)));
       else if (a.match(/^-[^-]+/) && !a.match(I)) {
         l = a.slice(1, -1).split(""), n = !1;
         for (let x = 0; x < l.length; x++) {
           if (d = a.slice(x + 2), l[x + 1] && l[x + 1] === "=") {
-            v = a.slice(x + 3), t = l[x], g(t, m.arrays) ? e = K(e, t, r, v) : g(t, m.nargs) !== !1 ? e = j(e, t, r, v) : b(t, v), n = !0;
+            v = a.slice(x + 3), t = l[x], g(t, m.arrays) ? e = K(e, t, r, v) : g(t, m.nargs) !== !1 ? e = j(e, t, r, v) : w(t, v), n = !0;
             break;
           }
           if (d === "-") {
-            b(l[x], d);
+            w(l[x], d);
             continue;
           }
           if (/[A-Za-z]/.test(l[x]) && /^-?\d+(\.\d*)?(e-?\d+)?$/.test(d) && g(d, m.bools) === !1) {
-            b(l[x], d), n = !0;
+            w(l[x], d), n = !0;
             break;
           }
           if (l[x + 1] && l[x + 1].match(/\W/)) {
-            b(l[x], d), n = !0;
+            w(l[x], d), n = !0;
             break;
           } else
-            b(l[x], R(l[x]));
+            w(l[x], R(l[x]));
         }
-        t = a.slice(-1)[0], !n && t !== "-" && (g(t, m.arrays) ? e = K(e, t, r) : g(t, m.nargs) !== !1 ? e = j(e, t, r) : (d = r[e + 1], d !== void 0 && (!/^(-|--)[^-]/.test(d) || d.match(I)) && !g(t, m.bools) && !g(t, m.counts) || /^(true|false)$/.test(d) ? (b(t, d), e++) : b(t, R(t))));
+        t = a.slice(-1)[0], !n && t !== "-" && (g(t, m.arrays) ? e = K(e, t, r) : g(t, m.nargs) !== !1 ? e = j(e, t, r) : (d = r[e + 1], d !== void 0 && (!/^(-|--)[^-]/.test(d) || d.match(I)) && !g(t, m.bools) && !g(t, m.counts) || /^(true|false)$/.test(d) ? (w(t, d), e++) : w(t, R(t))));
       } else if (a.match(/^-[0-9]$/) && a.match(I) && g(a.slice(1), m.bools))
-        t = a.slice(1), b(t, R(t));
+        t = a.slice(1), w(t, R(t));
       else if (a === "--") {
         V = r.slice(e + 1);
         break;
@@ -224,23 +224,23 @@ class De {
       } else
         Y(a);
     }
-    me(w, !0), me(w, !1), fe(w), be(), ue(w, m.aliases, q, !0), we(w), k["set-placeholder-key"] && ve(w), Object.keys(m.counts).forEach(function(e) {
-      N(w, e.split(".")) || b(e, 0);
-    }), Q && V.length && (w[y] = []), V.forEach(function(e) {
-      w[y].push(e);
-    }), k["camel-case-expansion"] && k["strip-dashed"] && Object.keys(w).filter((e) => e !== "--" && e.includes("-")).forEach((e) => {
-      delete w[e];
+    me(b, !0), me(b, !1), fe(b), we(), ue(b, m.aliases, q, !0), be(b), k["set-placeholder-key"] && ve(b), Object.keys(m.counts).forEach(function(e) {
+      N(b, e.split(".")) || w(e, 0);
+    }), Q && V.length && (b[y] = []), V.forEach(function(e) {
+      b[y].push(e);
+    }), k["camel-case-expansion"] && k["strip-dashed"] && Object.keys(b).filter((e) => e !== "--" && e.includes("-")).forEach((e) => {
+      delete b[e];
     }), k["strip-aliased"] && [].concat(...Object.keys(_).map((e) => _[e])).forEach((e) => {
-      k["camel-case-expansion"] && e.includes("-") && delete w[e.split(".").map((a) => W(a)).join(".")], delete w[e];
+      k["camel-case-expansion"] && e.includes("-") && delete b[e.split(".").map((a) => W(a)).join(".")], delete b[e];
     });
     function Y(e) {
       const a = H("_", e);
-      (typeof a == "string" || typeof a == "number") && w._.push(a);
+      (typeof a == "string" || typeof a == "number") && b._.push(a);
     }
     function j(e, a, c, n) {
       let t, l = g(a, m.nargs);
       if (l = typeof l != "number" || isNaN(l) ? 1 : l, l === 0)
-        return G(n) || (L = Error(M("Argument unexpected for: %s", a))), b(a, R(a)), e;
+        return G(n) || (L = Error(M("Argument unexpected for: %s", a))), w(a, R(a)), e;
       let p = G(n) ? 0 : 1;
       if (k["nargs-eats-options"])
         c.length - (e + 1) + p < l && (L = Error(M("Not enough arguments following: %s", a))), p = l;
@@ -250,8 +250,8 @@ class De {
         p < l && (L = Error(M("Not enough arguments following: %s", a)));
       }
       let d = Math.min(p, l);
-      for (!G(n) && d > 0 && (b(a, n), d--), t = e + 1; t < d + e + 1; t++)
-        b(a, c[t]);
+      for (!G(n) && d > 0 && (w(a, n), d--), t = e + 1; t < d + e + 1; t++)
+        w(a, c[t]);
       return e + d;
     }
     function K(e, a, c, n) {
@@ -269,9 +269,9 @@ class De {
         for (let d = e + 1; d < c.length && !(!k["greedy-arrays"] && t.length > 0 || p && typeof p == "number" && t.length >= p || (l = c[d], /^-/.test(l) && !I.test(l) && !$(l))); d++)
           e = d, t.push(J(a, l, h));
       }
-      return typeof p == "number" && (p && t.length < p || isNaN(p) && t.length === 0) && (L = Error(M("Not enough arguments following: %s", a))), b(a, t), e;
+      return typeof p == "number" && (p && t.length < p || isNaN(p) && t.length === 0) && (L = Error(M("Not enough arguments following: %s", a))), w(a, t), e;
     }
-    function b(e, a, c = h) {
+    function w(e, a, c = h) {
       if (/-/.test(e) && k["camel-case-expansion"]) {
         const l = e.split(".").map(function(p) {
           return W(p);
@@ -279,13 +279,13 @@ class De {
         re(e, l);
       }
       const n = J(e, a, c), t = e.split(".");
-      E(w, t, n), m.aliases[e] && m.aliases[e].forEach(function(l) {
+      E(b, t, n), m.aliases[e] && m.aliases[e].forEach(function(l) {
         const p = l.split(".");
-        E(w, p, n);
+        E(b, p, n);
       }), t.length > 1 && k["dot-notation"] && (m.aliases[t[0]] || []).forEach(function(l) {
         let p = l.split(".");
         const d = [].concat(t);
-        d.shift(), p = p.concat(d), (m.aliases[e] || []).includes(p.join(".")) || E(w, p, n);
+        d.shift(), p = p.concat(d), (m.aliases[e] || []).includes(p.join(".")) || E(b, p, n);
       }), g(e, m.normalize) && !g(e, m.arrays) && [e].concat(m.aliases[e] || []).forEach(function(p) {
         Object.defineProperty(se, p, {
           enumerable: !0,
@@ -340,10 +340,10 @@ class De {
     function Z(e, a) {
       Object.keys(e).forEach(function(c) {
         const n = e[c], t = a ? a + "." + c : c;
-        typeof n == "object" && n !== null && !Array.isArray(n) && k["dot-notation"] ? Z(n, t) : (!N(w, t.split(".")) || g(t, m.arrays) && k["combine-arrays"]) && b(t, n);
+        typeof n == "object" && n !== null && !Array.isArray(n) && k["dot-notation"] ? Z(n, t) : (!N(b, t.split(".")) || g(t, m.arrays) && k["combine-arrays"]) && w(t, n);
       });
     }
-    function be() {
+    function we() {
       typeof T < "u" && T.forEach(function(e) {
         Z(e);
       });
@@ -357,11 +357,11 @@ class De {
           const l = t.split("__").map(function(p, d) {
             return d === 0 && (p = p.substring(c.length)), W(p);
           });
-          (a && m.configs[l.join(".")] || !a) && !N(e, l) && b(l.join("."), n[t]);
+          (a && m.configs[l.join(".")] || !a) && !N(e, l) && w(l.join("."), n[t]);
         }
       });
     }
-    function we(e) {
+    function be(e) {
       let a;
       const c = /* @__PURE__ */ new Set();
       Object.keys(e).forEach(function(n) {
@@ -486,7 +486,7 @@ class De {
     }
     return {
       aliases: Object.assign({}, m.aliases),
-      argv: Object.assign(se, w),
+      argv: Object.assign(se, b),
       configuration: k,
       defaulted: Object.assign({}, O),
       error: L,
@@ -1327,7 +1327,14 @@ const ne = [
     pricing_urls: ["https://aws.amazon.com/bedrock/pricing/"],
     api_pattern: "https://bedrock-runtime\\.[a-z0-9-]+\\.amazonaws\\.com/",
     provider_match: {
-      contains: "bedrock"
+      or: [
+        {
+          contains: "bedrock"
+        },
+        {
+          contains: "amazon"
+        }
+      ]
     },
     extractors: [
       {
@@ -4251,6 +4258,21 @@ Compared to other leading proprietary and open-weights models Command A delivers
           }
         }
       },
+      {
+        id: "gemini-3.5-flash",
+        name: "Gemini 3.5 Flash",
+        description: "Google's most intelligent model built for speed, combining frontier intelligence with improved reasoning, coding, and multimodal understanding.",
+        match: {
+          starts_with: "gemini-3.5-flash"
+        },
+        context_window: 1e6,
+        price_comments: "See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers offer 50% discount on input/output.",
+        prices: {
+          input_mtok: 1.5,
+          cache_read_mtok: 0.15,
+          output_mtok: 9
+        }
+      },
       {
         id: "gemini-embedding-001",
         match: {

package/dist/index.cjs

@@ -1,6 +1,6 @@
 "use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});const Q=[{id:"anthropic",name:"Anthropic",pricing_urls:["https://www.anthropic.com/pricing#api"],api_pattern:"https://api\\.anthropic\\.com",model_match:{contains:"claude"},provider_match:{contains:"anthropic"},extractors:[{api_flavor:"default",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"cache_creation_input_tokens",dest:"input_tokens",required:!1},{path:"cache_read_input_tokens",dest:"input_tokens",required:!1},{path:"cache_creation_input_tokens",dest:"cache_write_tokens",required:!1},{path:"cache_read_input_tokens",dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:"cached_tokens",dest:"cache_read_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"claude-2",name:"Claude 2.0 / 2.1",description:`Claude 2 is Anthropic's previous generation model, offering reliable performance for various tasks. This includes Claude 2.0 and Claude 2.1.
-`,match:{or:[{starts_with:"claude-2"},{contains:"claude-v2"}]},context_window:2e5,prices:{input_mtok:8,output_mtok:24}},{id:"claude-3-5-haiku-latest",name:"Claude Haiku 3.5",description:"Fastest, most cost-effective model",match:{or:[{starts_with:"claude-3-5-haiku"},{starts_with:"claude-3.5-haiku"}]},context_window:2e5,prices:{input_mtok:.8,cache_write_mtok:1,cache_read_mtok:.08,output_mtok:4}},{id:"claude-3-5-sonnet",name:"Claude Sonnet 3.5",description:"Claude 3.5 Sonnet is an ideal balance of intelligence and speed for enterprise workloads. Maximum utility at a lower price, dependable, balanced for scaled deployments.",match:{or:[{starts_with:"claude-3-5-sonnet"},{starts_with:"claude-3.5-sonnet"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-7-sonnet-latest",name:"Claude Sonnet 3.7",description:"Claude 3.7 Sonnet is an advanced large language model with improved reasoning, coding, and problem-solving capabilities.",match:{or:[{starts_with:"claude-3-7-sonnet"},{starts_with:"claude-3.7-sonnet"},{starts_with:"claude-sonnet-3.7"},{starts_with:"claude-sonnet-3-7"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-haiku",name:"Claude Haiku 3",description:"Fastest, most cost-effective model",match:{starts_with:"claude-3-haiku"},context_window:2e5,prices:{input_mtok:.25,cache_write_mtok:.3,cache_read_mtok:.03,output_mtok:1.25}},{id:"claude-3-opus-latest",name:"Claude Opus 3",description:"Claude 3 Opus was Anthropic's most powerful model for highly complex tasks. It boasts top-level performance, intelligence, fluency, and understanding.",match:{starts_with:"claude-3-opus"},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-3-sonnet",name:"Claude 3 Sonnet",description:"Claude 3 Sonnet is an ideal balance of intelligence and speed for enterprise workloads. Maximum utility at a lower price, dependable, balanced for scaled deployments.",match:{starts_with:"claude-3-sonnet"},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-haiku-4-5",name:"Claude Haiku 4.5",description:"Fastest and most intelligent Haiku model",match:{or:[{starts_with:"claude-haiku-4-5"},{starts_with:"claude-haiku-4.5"},{starts_with:"claude-4-5-haiku"},{starts_with:"claude-4.5-haiku"}]},context_window:2e5,prices:{input_mtok:1,cache_write_mtok:1.25,cache_read_mtok:.1,output_mtok:5}},{id:"claude-opus-4-0",name:"Claude Opus 4",description:"Most intelligent model for complex tasks",match:{or:[{starts_with:"claude-opus-4-0"},{starts_with:"claude-4-opus"},{equals:"claude-opus-4"},{equals:"claude-opus-4-20250514"}]},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-opus-4-1",name:"Claude Opus 4.1",description:"Most intelligent model for complex tasks",match:{or:[{starts_with:"claude-opus-4-1"},{starts_with:"claude-opus-4.1"}]},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-opus-4-5",name:"Claude Opus 4.5",description:"Premium model combining maximum intelligence with practical performance",match:{or:[{starts_with:"claude-opus-4-5"},{starts_with:"claude-opus-4.5"},{starts_with:"claude-4-5-opus"},{starts_with:"claude-4.5-opus"}]},context_window:2e5,prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"claude-opus-4-6",name:"Claude Opus 4.6",description:"Our most intelligent model for building agents and coding",match:{or:[{starts_with:"claude-opus-4-6"},{starts_with:"claude-opus-4.6"},{starts_with:"claude-4-6-opus"},{starts_with:"claude-4.6-opus"}]},context_window:2e5,prices:[{prices:{input_mtok:{base:5,tiers:[{start:2e5,price:10}]},cache_write_mtok:{base:6.25,tiers:[{start:2e5,price:12.5}]},cache_read_mtok:{base:.5,tiers:[{start:2e5,price:1}]},output_mtok:{base:25,tiers:[{start:2e5,price:37.5}]}}},{constraint:{start_date:"2026-03-13",type:"start_date"},prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}}]},{id:"claude-opus-4-7",name:"Claude Opus 4.7",description:"Our most capable model for complex reasoning and agentic coding",match:{or:[{starts_with:"claude-opus-4-7"},{starts_with:"claude-opus-4.7"},{starts_with:"claude-4-7-opus"},{starts_with:"claude-4.7-opus"}]},context_window:1e6,price_comments:"Flat pricing across full 1M context window (no tiered pricing). Ref: https://platform.claude.com/docs/en/about-claude/pricing#long-context-pricing",prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"claude-sonnet-4-0",name:"Claude Sonnet 4",description:"Optimal balance of intelligence, cost, and speed",match:{or:[{starts_with:"claude-sonnet-4-2025"},{starts_with:"claude-sonnet-4-0"},{starts_with:"claude-sonnet-4@"},{equals:"claude-sonnet-4"},{starts_with:"claude-4-sonnet"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-sonnet-4-5",name:"Claude Sonnet 4.5",description:"Our best combination of speed and intelligence",match:{or:[{starts_with:"claude-sonnet-4-5"},{starts_with:"claude-sonnet-4.5"}]},context_window:1e6,prices:{input_mtok:{base:3,tiers:[{start:2e5,price:6}]},cache_write_mtok:{base:3.75,tiers:[{start:2e5,price:7.5}]},cache_read_mtok:{base:.3,tiers:[{start:2e5,price:.6}]},output_mtok:{base:15,tiers:[{start:2e5,price:22.5}]}}},{id:"claude-sonnet-4-6",name:"Claude Sonnet 4.6",description:"Our best combination of speed and intelligence",match:{or:[{starts_with:"claude-sonnet-4-6"},{starts_with:"claude-sonnet-4.6"}]},context_window:1e6,prices:[{prices:{input_mtok:{base:3,tiers:[{start:2e5,price:6}]},cache_write_mtok:{base:3.75,tiers:[{start:2e5,price:7.5}]},cache_read_mtok:{base:.3,tiers:[{start:2e5,price:.6}]},output_mtok:{base:15,tiers:[{start:2e5,price:22.5}]}}},{constraint:{start_date:"2026-03-13",type:"start_date"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}}]},{id:"claude-v1",description:"Retired, here to match price sources",match:{equals:"claude-v1"},prices:{input_mtok:8,output_mtok:24}}]},{id:"avian",name:"Avian",pricing_urls:["https://avian.io/pricing/"],api_pattern:"https://api\\.avian\\.io",models:[{id:"Meta-Llama-3.1-405B-Instruct",match:{equals:"Meta-Llama-3.1-405B-Instruct"},prices:{input_mtok:1.5,output_mtok:1.5}},{id:"Meta-Llama-3.1-70B-Instruct",match:{equals:"Meta-Llama-3.1-70B-Instruct"},prices:{input_mtok:.45,output_mtok:.45}},{id:"Meta-Llama-3.1-8B-Instruct",match:{equals:"Meta-Llama-3.1-8B-Instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"Meta-Llama-3.3-70B-Instruct",match:{equals:"Meta-Llama-3.3-70B-Instruct"},prices:{input_mtok:.45,output_mtok:.45}}]},{id:"aws",name:"AWS Bedrock",pricing_urls:["https://aws.amazon.com/bedrock/pricing/"],api_pattern:"https://bedrock-runtime\\.[a-z0-9-]+\\.amazonaws\\.com/",provider_match:{contains:"bedrock"},extractors:[{api_flavor:"default",root:"usage",model_path:"model",mappings:[{path:"inputTokens",dest:"input_tokens",required:!0},{path:"outputTokens",dest:"output_tokens",required:!0}]},{api_flavor:"anthropic",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"cache_creation_input_tokens",dest:"input_tokens",required:!1},{path:"cache_read_input_tokens",dest:"input_tokens",required:!1},{path:"cache_creation_input_tokens",dest:"cache_write_tokens",required:!1},{path:"cache_read_input_tokens",dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"amazon.nova-lite-v1:0",name:"Nova Lite",description:"Amazon Nova Lite 1.0 is a very low-cost multimodal model from Amazon that focused on fast processing of image, video, and text inputs to generate text output. Amazon Nova Lite can handle real-time customer interactions, document analysis, and visual question-answering tasks with high accuracy.",match:{contains:"amazon.nova-lite"},prices:{input_mtok:.06,cache_read_mtok:.015,output_mtok:.24}},{id:"amazon.nova-micro-v1:0",name:"Nova Micro",description:"Amazon Nova Micro 1.0 is a text-only model that delivers the lowest latency responses in the Amazon Nova family of models at a very low cost. With a context length of 128K tokens and optimized for speed and cost, Amazon Nova Micro excels at tasks such as text summarization, translation, content classification, interactive chat, and brainstorming. It has  simple mathematical reasoning and coding abilities.",match:{contains:"amazon.nova-micro"},prices:{input_mtok:.035,cache_read_mtok:.00875,output_mtok:.14}},{id:"amazon.nova-premier-v1:0",name:"Nova Premier",match:{contains:"amazon.nova-premier"},prices:{input_mtok:2.5,cache_read_mtok:.625,output_mtok:12.5}},{id:"amazon.nova-pro-v1:0",name:"Nova Pro",description:"Amazon Nova Pro 1.0 is a capable multimodal model from Amazon focused on providing a combination of accuracy, speed, and cost for a wide range of tasks. As of December 2024, it achieves state-of-the-art performance on key benchmarks including visual question answering (TextVQA) and video understanding (VATEX).",match:{contains:"amazon.nova-pro"},prices:{input_mtok:.8,cache_read_mtok:.2,output_mtok:3.2}},{id:"amazon.nova-sonic-v1:0",name:"Nova Sonic",match:{contains:"amazon.nova-sonic"},prices:{input_mtok:.06,output_mtok:.24,input_audio_mtok:3.4,output_audio_mtok:13.6}},{id:"amazon.titan-embed-text-v1",name:"Titan Embeddings G1 - Text",match:{contains:"amazon.titan-embed-text"},prices:{input_mtok:.1}},{id:"amazon.titan-text-express-v1",name:"Titan Text G1 - Express",match:{contains:"titan-text-express"},prices:{input_mtok:.2,output_mtok:.6}},{id:"amazon.titan-text-lite-v1",name:"Titan Text G1 - Lite",match:{contains:"titan-text-lite"},prices:{input_mtok:.15,output_mtok:.2}},{id:"deepseek.r1-v1:0",name:"DeepSeek-R1",match:{contains:"deepseek.r1"},prices:{input_mtok:1.35,output_mtok:5.4}},{id:"global.anthropic.claude-haiku-4-5-20251001-v1:0",match:{contains:"global.anthropic.claude-haiku-4-5-20251001"},prices:{input_mtok:1,cache_write_mtok:1.25,cache_read_mtok:.1,output_mtok:5}},{id:"global.anthropic.claude-opus-4-5-v1:0",match:{contains:"global.anthropic.claude-opus-4-5"},prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"global.anthropic.claude-opus-4-6-v1:0",match:{contains:"global.anthropic.claude-opus-4-6"},prices:{input_mtok:{base:5,tiers:[{start:2e5,price:10}]},cache_write_mtok:{base:6.25,tiers:[{start:2e5,price:12.5}]},cache_read_mtok:{base:.5,tiers:[{start:2e5,price:1}]},output_mtok:{base:25,tiers:[{start:2e5,price:37.5}]}}},{id:"global.anthropic.claude-opus-4-7-v1:0",match:{contains:"global.anthropic.claude-opus-4-7"},prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"global.anthropic.claude-sonnet-4-20250514-v1:0",match:{contains:"global.anthropic.claude-sonnet-4-20250514"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"global.anthropic.claude-sonnet-4-5-20250929-v1:0",match:{contains:"global.anthropic.claude-sonnet-4-5-20250929"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"global.anthropic.claude-sonnet-4-6-v1:0",match:{contains:"global.anthropic.claude-sonnet-4-6"},prices:{input_mtok:{base:3,tiers:[{start:2e5,price:6}]},cache_write_mtok:{base:3.75,tiers:[{start:2e5,price:7.5}]},cache_read_mtok:{base:.3,tiers:[{start:2e5,price:.6}]},output_mtok:{base:15,tiers:[{start:2e5,price:22.5}]}}},{id:"meta.llama3-1-70b-instruct-v1:0",name:"Llama 3.1 70B Instruct",match:{contains:"meta.llama3-1-70b-instruct"},prices:{input_mtok:.72,output_mtok:.72}},{id:"meta.llama3-1-8b-instruct-v1:0",name:"Llama 3.1 8B Instruct",match:{contains:"meta.llama3-1-8b-instruct"},prices:{input_mtok:.22,output_mtok:.22}},{id:"meta.llama3-2-11b-instruct-v1:0",name:"Llama 3.2 11B Instruct",match:{contains:"meta.llama3-2-11b-instruct"},prices:{input_mtok:.16,output_mtok:.16}},{id:"meta.llama3-2-1b-instruct-v1:0",name:"Llama 3.2 1B Instruct",match:{contains:"meta.llama3-2-1b-instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"meta.llama3-2-3b-instruct-v1:0",name:"Llama 3.2 3B Instruct",match:{contains:"meta.llama3-2-3b-instruct"},prices:{input_mtok:.15,output_mtok:.15}},{id:"meta.llama3-2-90b-instruct-v1:0",name:"Llama 3.2 90B Instruct",match:{contains:"meta.llama3-2-90b-instruct"},prices:{input_mtok:.72,output_mtok:.72}},{id:"meta.llama3-3-70b-instruct-v1:0",name:"Llama 3.3 70B Instruct",match:{contains:"meta.llama3-3-70b-instruct"},prices:{input_mtok:.72,output_mtok:.72}},{id:"meta.llama3-70b-instruct-v1:0",name:"Llama 3 70B Instruct",match:{contains:"meta.llama3-70b-instruct"},prices:{input_mtok:2.65,output_mtok:3.5}},{id:"meta.llama3-8b-instruct-v1:0",name:"Llama 3 8B Instruct",match:{contains:"meta.llama3-8b-instruct"},prices:{input_mtok:.3,output_mtok:.6}},{id:"meta.llama4-maverick-17b-instruct-v1:0",name:"Llama 4 Maverick 17B Instruct",match:{contains:"meta.llama4-maverick-17b-instruct"},prices:{input_mtok:.24,output_mtok:.97}},{id:"meta.llama4-scout-17b-instruct-v1:0",name:"Llama 4 Scout 17B Instruct",match:{contains:"meta.llama4-scout-17b-instruct"},prices:{input_mtok:.17,output_mtok:.66}},{id:"mistral.mistral-7b-instruct-v0:2",name:"Mistral 7B Instruct",match:{contains:"mistral.mistral-7b-instruct-v0"},prices:{input_mtok:.15,output_mtok:.2}},{id:"mistral.mistral-large-2402-v1:0",name:"Mistral Large (24.02)",match:{contains:"mistral.mistral-large-2402"},prices:{input_mtok:4,output_mtok:12}},{id:"mistral.mistral-small-2402-v1:0",name:"Mistral Small (24.02)",match:{contains:"mistral.mistral-small-2402"},prices:{input_mtok:1,output_mtok:3}},{id:"mistral.mixtral-8x7b-instruct-v0:1",name:"Mixtral 8x7B Instruct",match:{contains:"mistral.mixtral-8x7b-instruct-v0"},prices:{input_mtok:.45,output_mtok:.7}},{id:"mistral.pixtral-large-2502-v1:0",name:"Pixtral Large (25.02)",match:{contains:"mistral.pixtral-large-2502"},prices:{input_mtok:2,output_mtok:6}},{id:"nvidia.nemotron-nano-3-30b:0",name:"Nemotron 3 Nano 30B",match:{contains:"nvidia.nemotron-nano-3-30b"},prices:{input_mtok:.06,output_mtok:.24}},{id:"nvidia.nemotron-nano-9b-v2:0",name:"Nemotron 2 Nano 9B",match:{contains:"nvidia.nemotron-nano-9b-v2"},prices:{input_mtok:.06,output_mtok:.23}},{id:"nvidia.nemotron-super-3-120b:0",name:"Nemotron 3 Super 120B",match:{contains:"nvidia.nemotron-super-3-120b"},prices:{input_mtok:.15,output_mtok:.65}},{id:"openai.gpt-oss-120b-1:0",name:"gpt-oss-120b",match:{contains:"openai.gpt-oss-120b-1"},prices:{input_mtok:.15,output_mtok:.6}},{id:"openai.gpt-oss-20b-1:0",name:"gpt-oss-20b",match:{contains:"openai.gpt-oss-20b-1"},prices:{input_mtok:.07,output_mtok:.3}},{id:"qwen.qwen3-32b-v1:0",name:"Qwen3 32B (dense)",match:{contains:"qwen.qwen3-32b"},prices:{input_mtok:.15,output_mtok:.6}},{id:"qwen.qwen3-coder-30b-a3b-v1:0",name:"Qwen3-Coder-30B-A3B-Instruct",match:{contains:"qwen.qwen3-coder-30b-a3b"},prices:{input_mtok:.15,output_mtok:.6}},{id:"qwen.qwen3-coder-480b-a35b-v1:0",name:"Qwen3-Coder-480B-A35B-Instruct",match:{contains:"qwen.qwen3-coder-480b-a35b"},prices:{input_mtok:.45,output_mtok:1.8}},{id:"qwen.qwen3-vl-235b-a22b-v1:0",name:"Qwen3-VL-235B-A22B-Instruct",match:{contains:"qwen.qwen3-vl-235b-a22b"},prices:{input_mtok:.53,output_mtok:2.66}},{id:"regional.anthropic.claude-3-5-haiku-20241022-v1:0",match:{contains:"claude-3-5-haiku-20241022"},prices:{input_mtok:.8,cache_write_mtok:1,cache_read_mtok:.08,output_mtok:4}},{id:"regional.anthropic.claude-3-5-sonnet-20240620-v1:0",match:{contains:"claude-3-5-sonnet-20240620"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-3-5-sonnet-20241022-v2:0",match:{contains:"claude-3-5-sonnet-20241022"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-3-7-sonnet-20250219-v1:0",match:{contains:"claude-3-7-sonnet-20250219"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-3-haiku-20240307-v1:0",match:{contains:"claude-3-haiku-20240307"},prices:{input_mtok:.25,output_mtok:1.25}},{id:"regional.anthropic.claude-3-opus-20240229-v1:0",match:{contains:"claude-3-opus-20240229"},prices:{input_mtok:15,output_mtok:75}},{id:"regional.anthropic.claude-3-sonnet-20240229-v1:0",match:{contains:"claude-3-sonnet-20240229"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-haiku-4-5-20251001-v1:0",match:{or:[{starts_with:"anthropic.claude-haiku-4-5-20251001"},{starts_with:"claude-haiku-4-5-20251001"},{contains:"us.anthropic.claude-haiku-4-5-20251001"},{contains:"au.anthropic.claude-haiku-4-5-20251001"},{contains:"apac.anthropic.claude-haiku-4-5-20251001"},{contains:"eu.anthropic.claude-haiku-4-5-20251001"},{contains:"us-gov.anthropic.claude-haiku-4-5-20251001"},{contains:"jp.anthropic.claude-haiku-4-5-20251001"}]},prices:{input_mtok:1.1,cache_write_mtok:1.375,cache_read_mtok:.11,output_mtok:5.5}},{id:"regional.anthropic.claude-opus-4-1-20250805-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-1-20250805"},{starts_with:"claude-opus-4-1-20250805"},{contains:"us.anthropic.claude-opus-4-1-20250805"},{contains:"au.anthropic.claude-opus-4-1-20250805"},{contains:"apac.anthropic.claude-opus-4-1-20250805"},{contains:"eu.anthropic.claude-opus-4-1-20250805"},{contains:"us-gov.anthropic.claude-opus-4-1-20250805"},{contains:"jp.anthropic.claude-opus-4-1-20250805"}]},prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"regional.anthropic.claude-opus-4-20250514-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-20250514"},{starts_with:"claude-opus-4-20250514"},{contains:"us.anthropic.claude-opus-4-20250514"},{contains:"au.anthropic.claude-opus-4-20250514"},{contains:"apac.anthropic.claude-opus-4-20250514"},{contains:"eu.anthropic.claude-opus-4-20250514"},{contains:"us-gov.anthropic.claude-opus-4-20250514"},{contains:"jp.anthropic.claude-opus-4-20250514"}]},prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"regional.anthropic.claude-opus-4-5-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-5"},{starts_with:"claude-opus-4-5"},{contains:"us.anthropic.claude-opus-4-5"},{contains:"au.anthropic.claude-opus-4-5"},{contains:"apac.anthropic.claude-opus-4-5"},{contains:"eu.anthropic.claude-opus-4-5"},{contains:"us-gov.anthropic.claude-opus-4-5"},{contains:"jp.anthropic.claude-opus-4-5"}]},prices:{input_mtok:5.5,cache_write_mtok:6.875,cache_read_mtok:.55,output_mtok:27.5}},{id:"regional.anthropic.claude-opus-4-6-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-6"},{starts_with:"claude-opus-4-6"},{contains:"us.anthropic.claude-opus-4-6"},{contains:"au.anthropic.claude-opus-4-6"},{contains:"apac.anthropic.claude-opus-4-6"},{contains:"eu.anthropic.claude-opus-4-6"},{contains:"us-gov.anthropic.claude-opus-4-6"},{contains:"jp.anthropic.claude-opus-4-6"}]},prices:{input_mtok:{base:5.5,tiers:[{start:2e5,price:11}]},cache_write_mtok:{base:6.875,tiers:[{start:2e5,price:13.75}]},cache_read_mtok:{base:.55,tiers:[{start:2e5,price:1.1}]},output_mtok:{base:27.5,tiers:[{start:2e5,price:41.25}]}}},{id:"regional.anthropic.claude-opus-4-7-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-7"},{starts_with:"claude-opus-4-7"},{contains:"us.anthropic.claude-opus-4-7"},{contains:"au.anthropic.claude-opus-4-7"},{contains:"apac.anthropic.claude-opus-4-7"},{contains:"eu.anthropic.claude-opus-4-7"},{contains:"us-gov.anthropic.claude-opus-4-7"},{contains:"jp.anthropic.claude-opus-4-7"}]},prices:{input_mtok:5.5,cache_write_mtok:6.875,cache_read_mtok:.55,output_mtok:27.5}},{id:"regional.anthropic.claude-sonnet-4-20250514-v1:0",match:{or:[{starts_with:"anthropic.claude-sonnet-4-20250514"},{starts_with:"claude-sonnet-4-20250514"},{contains:"us.anthropic.claude-sonnet-4-20250514"},{contains:"au.anthropic.claude-sonnet-4-20250514"},{contains:"apac.anthropic.claude-sonnet-4-20250514"},{contains:"eu.anthropic.claude-sonnet-4-20250514"},{contains:"us-gov.anthropic.claude-sonnet-4-20250514"},{contains:"jp.anthropic.claude-sonnet-4-20250514"}]},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-sonnet-4-5-20250929-v1:0",match:{or:[{starts_with:"anthropic.claude-sonnet-4-5-20250929"},{starts_with:"claude-sonnet-4-5-20250929"},{contains:"us.anthropic.claude-sonnet-4-5-20250929"},{contains:"au.anthropic.claude-sonnet-4-5-20250929"},{contains:"apac.anthropic.claude-sonnet-4-5-20250929"},{contains:"eu.anthropic.claude-sonnet-4-5-20250929"},{contains:"us-gov.anthropic.claude-sonnet-4-5-20250929"},{contains:"jp.anthropic.claude-sonnet-4-5-20250929"}]},prices:{input_mtok:3.3,cache_write_mtok:4.125,cache_read_mtok:.33,output_mtok:16.5}},{id:"regional.anthropic.claude-sonnet-4-6-v1:0",match:{or:[{starts_with:"anthropic.claude-sonnet-4-6"},{starts_with:"claude-sonnet-4-6"},{contains:"us.anthropic.claude-sonnet-4-6"},{contains:"au.anthropic.claude-sonnet-4-6"},{contains:"apac.anthropic.claude-sonnet-4-6"},{contains:"eu.anthropic.claude-sonnet-4-6"},{contains:"us-gov.anthropic.claude-sonnet-4-6"},{contains:"jp.anthropic.claude-sonnet-4-6"}]},prices:{input_mtok:{base:3.3,tiers:[{start:2e5,price:6.6}]},cache_write_mtok:{base:4.125,tiers:[{start:2e5,price:8.25}]},cache_read_mtok:{base:.33,tiers:[{start:2e5,price:.66}]},output_mtok:{base:16.5,tiers:[{start:2e5,price:24.75}]}}}]},{id:"azure",name:"Microsoft Azure",pricing_urls:["https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/#pricing"],api_pattern:"(https?://)?([^.]*\\.)?(?:openai\\.azure\\.com|azure-api\\.net|cognitiveservices\\.azure\\.com)",price_comments:'These are prices for "*-Global" models, prices for "Regional" models are often slightly higher. Retired models are listed at https://learn.microsoft.com/th-th/azure/ai-foundry/openai/concepts/legacy-models',extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["prompt_tokens_details","audio_tokens"],dest:"input_audio_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"responses",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:["input_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"embeddings",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0}]},{api_flavor:"anthropic",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"cache_creation_input_tokens",dest:"input_tokens",required:!1},{path:"cache_read_input_tokens",dest:"input_tokens",required:!1},{path:"cache_creation_input_tokens",dest:"cache_write_tokens",required:!1},{path:"cache_read_input_tokens",dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]}],fallback_model_providers:["openai","anthropic"],models:[{id:"ada",match:{or:[{equals:"ada"},{equals:"text-embedding-ada"},{equals:"text-embedding-ada-002"},{equals:"text-embedding-ada-002-v2"}]},prices:{input_mtok:.1}},{id:"babbage",match:{or:[{equals:"babbage"},{equals:"babbage-002"}]},prices:{input_mtok:.4}},{id:"curie",match:{or:[{equals:"curie"},{equals:"text-curie"},{equals:"text-curie-001"}]},prices:{input_mtok:2}},{id:"davinci",match:{or:[{equals:"davinci"},{equals:"davinci-002"},{equals:"text-davinci"},{equals:"text-davinci-002"}]},prices:{input_mtok:2}},{id:"mai-ds-r1:free",name:"MAI DS R1 (free)",description:"MAI-DS-R1 is a post-trained variant of DeepSeek-R1 developed by the Microsoft AI team to improve the model's responsiveness on previously blocked topics while enhancing its safety profile. Built on top of DeepSeek-R1's reasoning foundation, it integrates 110k examples from the Tulu-3 SFT dataset and 350k internally curated multilingual safety-alignment samples. The model retains strong reasoning, coding, and problem-solving capabilities, while unblocking a wide range of prompts previously restricted in R1.",match:{equals:"mai-ds-r1:free"},prices:{}},{id:"o1",match:{or:[{equals:"o1"},{equals:"o1-2024-12-17"},{equals:"o1-preview"},{equals:"o1-preview-2024-09-12"}]},prices:{input_mtok:15,cache_read_mtok:7.5,output_mtok:60}},{id:"o1-mini",match:{or:[{equals:"o1-mini"},{equals:"o1-mini-2024-09-12"}]},prices:{input_mtok:1.1,cache_read_mtok:.55,output_mtok:4.4}},{id:"o3-2025-04-16",match:{or:[{equals:"o3"},{equals:"o3-2025-04-16"}]},prices:{input_mtok:2,cache_read_mtok:.5,output_mtok:8}},{id:"o3-mini",match:{or:[{equals:"o3-mini"},{equals:"o3-mini-2025-01-31"}]},prices:{input_mtok:1.1,cache_read_mtok:.55,output_mtok:4.4}},{id:"o4-mini",match:{or:[{contains:"o4-mini"},{contains:"o4-mini-2025-04-16"}]},prices:{input_mtok:1.1,cache_read_mtok:.28,output_mtok:4.4}},{id:"phi-3-medium-128k-instruct",name:"Phi-3 Medium 128K Instruct",description:"Phi-3 128K Medium is a powerful 14-billion parameter model designed for advanced language understanding, reasoning, and instruction following. Optimized through supervised fine-tuning and preference adjustments, it excels in tasks involving common sense, mathematics, logical reasoning, and code processing.",match:{equals:"phi-3-medium-128k-instruct"},prices:{input_mtok:1,output_mtok:1}},{id:"phi-3-mini-128k-instruct",name:"Phi-3 Mini 128K Instruct",description:"Phi-3 Mini is a powerful 3.8B parameter model designed for advanced language understanding, reasoning, and instruction following. Optimized through supervised fine-tuning and preference adjustments, it excels in tasks involving common sense, mathematics, logical reasoning, and code processing.",match:{equals:"phi-3-mini-128k-instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"phi-3.5-mini-128k-instruct",name:"Phi-3.5 Mini 128K Instruct",description:"Phi-3.5 models are lightweight, state-of-the-art open models. These models were trained with Phi-3 datasets that include both synthetic data and the filtered, publicly available websites data, with a focus on high quality and reasoning-dense properties. Phi-3.5 Mini uses 3.8B parameters, and is a dense decoder-only transformer model using the same tokenizer as Phi-3 Mini.",match:{equals:"phi-3.5-mini-128k-instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"phi-4",name:"Phi 4",description:"Microsoft Research Phi-4 is designed to perform well in complex reasoning tasks and can operate efficiently in situations with limited memory or where quick responses are needed.",match:{equals:"phi-4"},prices:{input_mtok:.07,output_mtok:.14}},{id:"phi-4-multimodal-instruct",name:"Phi 4 Multimodal Instruct",description:"Phi-4 Multimodal Instruct is a versatile 5.6B parameter foundation model that combines advanced reasoning and instruction-following capabilities across both text and visual inputs, providing accurate text outputs. The unified architecture enables efficient, low-latency inference, suitable for edge and mobile deployments. Phi-4 Multimodal Instruct supports text inputs in multiple languages including Arabic, Chinese, English, French, German, Japanese, Spanish, and more, with visual input optimized primarily for English. It delivers impressive performance on multimodal tasks involving mathematical, scientific, and document reasoning, providing developers and enterprises a powerful yet compact model for sophisticated interactive applications. For more information, see the Phi-4 Multimodal blog post.",match:{equals:"phi-4-multimodal-instruct"},prices:{input_mtok:.05,output_mtok:.1}},{id:"phi-4-reasoning-plus",name:"Phi 4 Reasoning Plus",description:"Phi-4-reasoning-plus is an enhanced 14B parameter model from Microsoft, fine-tuned from Phi-4 with additional reinforcement learning to boost accuracy on math, science, and code reasoning tasks. It uses the same dense decoder-only transformer architecture as Phi-4, but generates longer, more comprehensive outputs structured into a step-by-step reasoning trace and final answer.",match:{equals:"phi-4-reasoning-plus"},prices:{input_mtok:.07,output_mtok:.35}},{id:"phi-4-reasoning-plus:free",name:"Phi 4 Reasoning Plus (free)",description:"Phi-4-reasoning-plus is an enhanced 14B parameter model from Microsoft, fine-tuned from Phi-4 with additional reinforcement learning to boost accuracy on math, science, and code reasoning tasks. It uses the same dense decoder-only transformer architecture as Phi-4, but generates longer, more comprehensive outputs structured into a step-by-step reasoning trace and final answer.",match:{equals:"phi-4-reasoning-plus:free"},prices:{}},{id:"phi-4-reasoning:free",name:"Phi 4 Reasoning (free)",description:"Phi-4-reasoning is a 14B parameter dense decoder-only transformer developed by Microsoft, fine-tuned from Phi-4 to enhance complex reasoning capabilities. It uses a combination of supervised fine-tuning on chain-of-thought traces and reinforcement learning, targeting math, science, and code reasoning tasks. With a 32k context window and high inference efficiency, it is optimized for structured responses in a two-part format: reasoning trace followed by a final solution.",match:{equals:"phi-4-reasoning:free"},prices:{}},{id:"text-embedding-3-large",match:{equals:"text-embedding-3-large"},prices:{input_mtok:.13}},{id:"text-embedding-3-small",match:{equals:"text-embedding-3-small"},prices:{input_mtok:.02}},{id:"wizardlm-2-8x22b",name:"WizardLM-2 8x22B",description:"WizardLM-2 8x22B is Microsoft AI's most advanced Wizard model. It demonstrates highly competitive performance compared to leading proprietary models, and it consistently outperforms all existing state-of-the-art opensource models.",match:{equals:"wizardlm-2-8x22b"},prices:{input_mtok:.48,output_mtok:.48}}]},{id:"cerebras",name:"Cerebras",pricing_urls:["https://www.cerebras.ai/pricing#pricing","https://inference-docs.cerebras.ai/models/openai-oss"],api_pattern:"https://api\\.cerebras\\.ai",model_match:{contains:"cerebras"},provider_match:{contains:"cerebras"},extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"gpt-oss-120b",name:"GPT-OSS 120B",description:"OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with 120 billion parameters and 128 experts. Delivers frontier reasoning capabilities with record-breaking inference speeds on Cerebras hardware (~3,000 tokens/second).",match:{or:[{equals:"gpt-oss-120b"},{starts_with:"cerebras/gpt-oss-120b"},{starts_with:"cerebras:gpt-oss-120b"}]},context_window:131072,price_comments:"Developer tier pricing. Free tier: 65k context, Paid tier: 131k context.",prices:{input_mtok:.35,output_mtok:.75}},{id:"llama-3.3-70b",name:"Llama 3.3 70B",description:"Meta's enhanced 70B model delivering 405B-level accuracy. Optimized for chat, coding, instruction following, mathematics, and reasoning with high-speed inference on Cerebras hardware (~2,100 tokens/second).",match:{or:[{equals:"llama-3.3-70b"},{starts_with:"cerebras/llama-3.3-70b"},{starts_with:"cerebras:llama-3.3-70b"}]},context_window:128e3,price_comments:"Developer tier pricing. Free tier: 65k context, Paid tier: 128k context.",prices:{input_mtok:.85,output_mtok:1.2}},{id:"llama3.1-8b",name:"Llama 3.1 8B",description:"Meta's Llama 3.1 8B model for general-purpose tasks including chat, coding, and instruction following. Optimized for fast inference on Cerebras hardware (~2,200 tokens/second).",match:{or:[{equals:"llama3.1-8b"},{starts_with:"cerebras/llama3.1-8b"},{starts_with:"cerebras:llama3.1-8b"}]},context_window:32768,price_comments:"Developer tier pricing. Free tier: 8k context, Paid tier: 32k context.",prices:{input_mtok:.1,output_mtok:.1}},{id:"qwen-3-32b",name:"Qwen 3 32B",description:"Qwen's 32B parameter model with enhanced reasoning and coding capabilities. Supports both standard and reasoning modes for complex tasks, with fast inference speeds on Cerebras hardware (~2,600 tokens/second).",match:{or:[{equals:"qwen-3-32b"},{starts_with:"cerebras/qwen-3-32b"},{starts_with:"cerebras:qwen-3-32b"}]},context_window:131072,price_comments:"Developer tier pricing. Free tier: 65k context, Paid tier: 131k context.",prices:{input_mtok:.4,output_mtok:.8}},{id:"qwen-3-coder-480b",name:"qwen-3-coder-480b",match:{equals:"qwen-3-coder-480b"},price_comments:"Seems to be no longer available on cerebras, here to help with tests",prices:{}}]},{id:"cohere",name:"Cohere",pricing_urls:["https://cohere.com/pricing"],api_pattern:"https://api\\.cohere\\.ai",model_match:{starts_with:"command-"},provider_match:{contains:"cohere"},extractors:[{api_flavor:"default",root:["usage","billed_units"],model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"output_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"embeddings",root:["meta","billed_units"],model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0}]}],models:[{id:"command",name:"Command",description:"Command is an instruction-following conversational model that performs language tasks with high quality, more reliably and with a longer context than our base generative models.",match:{equals:"command"},prices:{input_mtok:1,output_mtok:2}},{id:"command-a",name:"Command A",description:`Command A is an open-weights 111B parameter model with a 256k context window focused on delivering great performance across agentic, multilingual, and coding use cases.
-Compared to other leading proprietary and open-weights models Command A delivers maximum performance with minimum hardware costs, excelling on business-critical agentic and multilingual tasks.`,match:{starts_with:"command-a"},prices:{input_mtok:2.5,output_mtok:10}},{id:"command-r",name:"Command R",description:"Command-R is a 35B parameter model that performs conversational language tasks at a higher quality, more reliably, and with a longer context than previous models. It can be used for complex workflows like code generation, retrieval augmented generation (RAG), tool use, and agents.",match:{or:[{equals:"command-r"},{equals:"command-r-08-2024"}]},prices:{input_mtok:.15,output_mtok:.6}},{id:"command-r-plus",name:"Command R+",description:"Command R+ is a new, 104B-parameter LLM from Cohere. It's useful for roleplay, general consumer usecases, and Retrieval Augmented Generation (RAG).",match:{or:[{equals:"command-r-plus"},{equals:"command-r-plus-08-2024"}]},prices:{input_mtok:2.5,output_mtok:10}},{id:"command-r7b",name:"Command R7B",description:"Command R7B (12-2024) is a small, fast update of the Command R+ model, delivered in December 2024. It excels at RAG, tool use, agents, and similar tasks requiring complex reasoning and multiple steps.",match:{or:[{equals:"command-r7b"},{equals:"command-r7b-12-2024"}]},prices:{input_mtok:.0375,output_mtok:.15}},{id:"embed-v4.0",name:"Embed v4.0",description:"Embed v4.0 is a state-of-the-art embedding model designed for precise retrieval across noisy, multilingual, and multimodal data.",match:{equals:"embed-v4.0"},context_window:128e3,prices:{input_mtok:.12}}]},{id:"deepseek",name:"Deepseek",pricing_urls:["https://api-docs.deepseek.com/quick_start/pricing"],api_pattern:"https://api\\.deepseek\\.com",price_comments:'Deepseek off-peak pricing applies "UTC 16:30-00:30" so we switch it around and use the off-peak pricing as the default (first) price then the second price with a constraint is the "standard" pricing that applies "UTC 00:30-16:30".',model_match:{contains:"deepseek"},extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"deepseek-chat",name:"DeepSeek Chat",description:"DeepSeek-V3 is the latest model from the DeepSeek team, building upon the instruction following and coding abilities of the previous versions. Pre-trained on nearly 15 trillion tokens, the reported evaluations reveal that the model outperforms other open-source models and rivals leading closed-source models.",match:{or:[{starts_with:"deepseek-chat"},{equals:"deepseek-chat-v3-0324"}]},context_window:64e3,prices:[{prices:{input_mtok:.135,cache_read_mtok:.035,output_mtok:.55}},{constraint:{start_time:"00:30:00Z",end_time:"16:30:00Z",type:"time_of_date"},prices:{input_mtok:.27,cache_read_mtok:.07,output_mtok:1.1}}]},{id:"deepseek-reasoner",name:"Deepseek R1",description:"DeepSeek R1 is here: Performance on par with OpenAI o1, but open-sourced and with fully open reasoning tokens. It's 671B parameters in size, with 37B active in an inference pass.",match:{or:[{equals:"deepseek-reasoner"},{starts_with:"deepseek-r1"},{equals:"deepseek-r1-0528"}]},context_window:64e3,prices:[{prices:{input_mtok:.135,cache_read_mtok:.035,output_mtok:.55}},{constraint:{start_time:"00:30:00Z",end_time:"16:30:00Z",type:"time_of_date"},prices:{input_mtok:.55,cache_read_mtok:.14,output_mtok:2.19}}]},{id:"deepseek-v4-flash",name:"DeepSeek V4 Flash",description:"DeepSeek-V4-Flash. Supports both non-thinking and thinking (default) modes, JSON output, tool calls, chat prefix completion, and FIM completion (non-thinking only).",match:{or:[{starts_with:"deepseek-v4-flash"}]},context_window:1e6,prices:{input_mtok:.14,cache_read_mtok:.0028,output_mtok:.28}},{id:"deepseek-v4-pro",name:"DeepSeek V4 Pro",description:"DeepSeek-V4-Pro. Supports both non-thinking and thinking (default) modes, JSON output, tool calls, chat prefix completion, and FIM completion (non-thinking only).",match:{or:[{starts_with:"deepseek-v4-pro"}]},context_window:1e6,price_comments:"Standard (non-promotional) pricing. DeepSeek is offering a temporary 75% promotional discount that is not reflected here.",prices:{input_mtok:1.74,cache_read_mtok:.0145,output_mtok:3.48}}]},{id:"fireworks",name:"Fireworks",pricing_urls:["https://fireworks.ai/pricing"],api_pattern:"https://api\\.fireworks\\.ai",model_match:{starts_with:"accounts/fireworks/models/"},extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"deepseek-r1-0528",name:"DeepSeek R1 0528",description:"The updated DeepSeek-R1-0528 model delivers major improvements in reasoning, inference, and accuracy through enhanced post-training optimization and greater computational resources. It now performs at a level approaching top-tier models like O3 and Gemini 2.5 Pro, with notable gains in complex tasks such as math and programming.",match:{equals:"accounts/fireworks/models/deepseek-r1-0528"},context_window:16e4,prices:{input_mtok:3,output_mtok:8}},{id:"deepseek-v3-0324",name:"Deepseek V3 03-24",description:"A strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token from Deepseek. Updated checkpoint.",match:{equals:"accounts/fireworks/models/deepseek-v3-0324"},context_window:16e4,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.9,output_mtok:.9}},{id:"deepseek-v3p2",name:"Deepseek V3.2",description:"Model from Deepseek that harmonizes high computational efficiency with superior reasoning and agent performance. 675B parameter MoE model.",match:{equals:"accounts/fireworks/models/deepseek-v3p2"},context_window:163840,prices:{input_mtok:.56,cache_read_mtok:.28,output_mtok:1.68}},{id:"gemma-3-27b-it",name:"Gemma 3 27B Instruct",match:{equals:"accounts/fireworks/models/gemma-3-27b-it"},context_window:131e3,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.1,output_mtok:.1}},{id:"glm-4p7",name:"GLM-4.7",description:"Next-generation general-purpose model from Z.ai optimized for coding, reasoning, and agentic workflows. 352B parameter MoE model with advanced thinking controls.",match:{equals:"accounts/fireworks/models/glm-4p7"},context_window:202752,prices:{input_mtok:.6,output_mtok:2.2}},{id:"gpt-oss-120b",name:"OpenAI gpt-oss-120b",description:"OpenAI's open-weight 117B parameter MoE model designed for production, general purpose, high reasoning use-cases. Features powerful reasoning, agentic tasks, and versatile developer use cases.",match:{equals:"accounts/fireworks/models/gpt-oss-120b"},context_window:131072,prices:{input_mtok:.15,cache_read_mtok:.07,output_mtok:.6}},{id:"gpt-oss-20b",name:"OpenAI gpt-oss-20b",description:"OpenAI's open-weight 21.5B parameter model designed for powerful reasoning, agentic tasks, and versatile developer use cases. Optimized for lower latency and local or specialized tasks.",match:{equals:"accounts/fireworks/models/gpt-oss-20b"},context_window:131072,prices:{input_mtok:.07,cache_read_mtok:.04,output_mtok:.3}},{id:"kimi-k2p5",name:"Kimi K2.5",description:"Moonshot AI's flagship agentic model. Unifies vision and text, thinking and non-thinking modes, and single-agent and multi-agent execution into one model. 1T parameter MoE model.",match:{equals:"accounts/fireworks/models/kimi-k2p5"},context_window:262144,prices:{input_mtok:.6,cache_read_mtok:.1,output_mtok:3}},{id:"llama-v3p1-8b-instruct",name:"Llama 3.1 8B Instruct",description:"The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes. The Llama 3.1 instruction tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.",match:{equals:"accounts/fireworks/models/llama-v3p1-8b-instruct"},context_window:131e3,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.2,output_mtok:.2}},{id:"llama4-maverick-instruct-basic",name:"Llama 4 Maverick Instruct (Basic)",description:"The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes. The Llama 3.1 instruction tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.",match:{equals:"accounts/fireworks/models/llama4-maverick-instruct-basic"},context_window:1e6,prices:{input_mtok:.22,output_mtok:.88}},{id:"minimax-m2p1",name:"MiniMax-M2.1",description:"Built for strong real-world performance across complex, multi-language, and agent-driven workflows. 228B parameter model with robust support for systems, backend, web, mobile, and office-style tasks.",match:{equals:"accounts/fireworks/models/minimax-m2p1"},context_window:204800,prices:{input_mtok:.3,output_mtok:1.2}},{id:"qwen2p5-vl-72b-instruct",name:"Qwen2.5-VL 72B Instruct",description:"Latest Qwen's VLM model",match:{equals:"accounts/fireworks/models/qwen2p5-vl-72b-instruct"},context_window:128e3,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.9,output_mtok:.9}},{id:"qwen3-235b-a22b",name:"Qwen3 235B-A22B",description:'Qwen3 is the latest evolution in the Qwen LLM series, featuring both dense and MoE models with major advancements in reasoning, agent capabilities, multilingual support, and instruction following. It uniquely allows seamless switching between "thinking" (for complex logic, math, coding) and "non-thinking" modes (for fast, general dialogue), delivering strong performance across tasks.',match:{equals:"accounts/fireworks/models/qwen3-235b-a22b"},context_window:128e3,prices:{input_mtok:.22,output_mtok:.88}}]},{id:"google",name:"Google",pricing_urls:["https://ai.google.dev/gemini-api/docs/pricing","https://cloud.google.com/vertex-ai/generative-ai/pricing"],api_pattern:"https://(.*\\.)?googleapis\\.com",model_match:{contains:"gemini"},provider_match:{or:[{contains:"google"},{contains:"vertex"},{contains:"gemini"}]},extractors:[{api_flavor:"default",root:"usageMetadata",model_path:"modelVersion",mappings:[{path:"promptTokenCount",dest:"input_tokens",required:!1},{path:"cachedContentTokenCount",dest:"cache_read_tokens",required:!1},{path:["cacheTokensDetails",{type:"array-match",field:"modality",match:{equals:"AUDIO"}},"tokenCount"],dest:"cache_audio_read_tokens",required:!1},{path:["promptTokensDetails",{type:"array-match",field:"modality",match:{equals:"AUDIO"}},"tokenCount"],dest:"input_audio_tokens",required:!1},{path:["candidatesTokensDetails",{type:"array-match",field:"modality",match:{equals:"AUDIO"}},"tokenCount"],dest:"output_audio_tokens",required:!1},{path:"candidatesTokenCount",dest:"output_tokens",required:!1},{path:"thoughtsTokenCount",dest:"output_tokens",required:!1},{path:"toolUsePromptTokenCount",dest:"input_tokens",required:!1}]},{api_flavor:"anthropic",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"cache_creation_input_tokens",dest:"input_tokens",required:!1},{path:"cache_read_input_tokens",dest:"input_tokens",required:!1},{path:"cache_creation_input_tokens",dest:"cache_write_tokens",required:!1},{path:"cache_read_input_tokens",dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["prompt_tokens_details","audio_tokens"],dest:"input_audio_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],fallback_model_providers:["anthropic"],models:[{id:"claude-3-5-haiku",match:{contains:"claude-3-5-haiku"},context_window:2e5,prices:{input_mtok:.8,cache_write_mtok:1,cache_read_mtok:.08,output_mtok:4}},{id:"claude-3-5-sonnet",match:{contains:"claude-3-5-sonnet"},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-7-sonnet",match:{contains:"claude-3-7-sonnet"},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-haiku",match:{contains:"claude-3-haiku"},context_window:2e5,prices:{input_mtok:.25,cache_write_mtok:.3,cache_read_mtok:.03,output_mtok:1.25}},{id:"claude-3-opus",match:{contains:"claude-3-opus"},prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-4-opus",match:{or:[{contains:"claude-4-opus"},{contains:"claude-opus-4@"},{contains:"claude-opus-4-0"},{contains:"claude-opus-4-1"},{equals:"claude-opus-4"}]},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-4-sonnet",match:{or:[{contains:"claude-4-sonnet"},{contains:"claude-sonnet-4"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-opus-4-6",match:{or:[{contains:"claude-4-6-opus"},{contains:"claude-opus-4-6"},{contains:"claude-4.6-opus"},{contains:"claude-opus-4.6"}]},context_window:2e5,prices:{input_mtok:{base:5,tiers:[{start:2e5,price:10}]},cache_write_mtok:{base:6.25,tiers:[{start:2e5,price:12.5}]},cache_read_mtok:{base:.5,tiers:[{start:2e5,price:1}]},output_mtok:{base:25,tiers:[{start:2e5,price:37.5}]}}},{id:"claude-opus-4-7",match:{or:[{contains:"claude-4-7-opus"},{contains:"claude-opus-4-7"},{contains:"claude-4.7-opus"},{contains:"claude-opus-4.7"}]},context_window:1e6,price_comments:"Flat pricing across full 1M context window. Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models",prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"gemini-1.0-pro-vision-001",name:"gemini 1.0 pro vision",description:"Google's first-generation advanced multimodal model that can understand text, code, and images. It provides strong reasoning capabilities and follows instructions effectively.",match:{equals:"gemini-1.0-pro-vision-001"},context_window:32768,price_comments:"I can't find anything about this model or it's pricing, so trusting the original source",prices:{input_mtok:.125,output_mtok:.375}},{id:"gemini-1.5-flash",name:"gemini 1.5 flash",description:"A faster, more cost-effective variant of Gemini 1.5 that maintains strong capabilities while optimizing for performance and cost efficiency. Suitable for production deployments requiring high throughput.",match:{contains:"gemini-1.5-flash"},context_window:1e6,prices:{input_mtok:{base:.075,tiers:[{start:128e3,price:.15}]},cache_read_mtok:{base:.01875,tiers:[{start:128e3,price:.0375}]},output_mtok:{base:.3,tiers:[{start:128e3,price:.6}]}}},{id:"gemini-1.5-pro",name:"gemini 1.5 Pro",description:"Google's most capable multimodal model with an extremely long context window of up to 1 million tokens. It excels at complex reasoning, long-form content processing, and multimodal understanding.",match:{contains:"gemini-1.5-pro"},context_window:1e6,prices:{input_mtok:{base:1.25,tiers:[{start:128e3,price:2.5}]},output_mtok:{base:5,tiers:[{start:128e3,price:10}]}}},{id:"gemini-2.0-flash",name:"gemini 2.0 flash",description:"The newest generation of Google's Gemini models, featuring improved reasoning, instruction following, and factual accuracy, with the Flash variant optimized for cost-efficiency and performance.",match:{or:[{ends_with:"gemini-2.0-flash"},{contains:"gemini-2.0-flash-0"},{contains:"gemini-2.0-flash-exp"},{contains:"gemini-2.0-flash-thinking"},{contains:"gemini-2.0-flash-latest"}]},context_window:1e6,prices:{input_mtok:.1,cache_read_mtok:.025,output_mtok:.4,input_audio_mtok:.7,cache_audio_read_mtok:.175}},{id:"gemini-2.0-flash-lite",name:"gemini 2.0 flash lite",description:"A lighter, more cost-effective version of Gemini 2.0 Flash, designed for applications requiring high efficiency while maintaining good performance. Ideal for high-volume, cost-sensitive deployments.",match:{contains:"gemini-2.0-flash-lite"},context_window:1e6,prices:{input_mtok:.075,output_mtok:.3}},{id:"gemini-2.5-flash",name:"Gemini 2.5 Flash",description:`Gemini 2.5 Flash is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking" capabilities, enabling it to provide responses with greater accuracy and nuanced context handling.`,match:{or:[{equals:"gemini-2.5-flash"},{equals:"gemini-2.5-flash-latest"},{equals:"gemini-2.5-flash-preview-09-2025"}]},prices:{input_mtok:.3,cache_read_mtok:.03,output_mtok:2.5,input_audio_mtok:1,cache_audio_read_mtok:.1}},{id:"gemini-2.5-flash-image",name:"Gemini 2.5 Flash Image",description:"Google's specialized image generation model optimized for fast, high-quality image generation. Outputs images at 1024x1024 resolution, with each image consuming 1290 output tokens.",match:{or:[{equals:"gemini-2.5-flash-image"},{equals:"gemini-2.5-flash-image-preview"}]},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-flash-image. Image output is priced at $30 per 1M tokens, with each 1024x1024 image = 1290 tokens = $0.039/image. Cache pricing is not available for this model.",prices:{input_mtok:.3,output_mtok:30}},{id:"gemini-2.5-flash-lite",name:"Gemini 2.5 Flash Lite",description:'Gemini 2.5 Flash-Lite is a lightweight reasoning model in the Gemini 2.5 family, optimized for ultra-low latency and cost efficiency. It offers improved throughput, faster token generation, and better performance across common benchmarks compared to earlier Flash models. By default, "thinking" (i.e. multi-pass reasoning) is disabled to prioritize speed, but developers can enable it via the Reasoning API parameter to selectively trade off cost for intelligence.',match:{or:[{equals:"gemini-2.5-flash-lite"},{starts_with:"gemini-2.5-flash-lite-preview"}]},context_window:1e6,prices:{input_mtok:.1,cache_read_mtok:.01,output_mtok:.4,input_audio_mtok:.3,cache_audio_read_mtok:.03}},{id:"gemini-2.5-flash-preview",name:"Gemini 2.5 Flash Preview 05-20",description:`Gemini 2.5 Flash May 20th Checkpoint is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking" capabilities, enabling it to provide responses with greater accuracy and nuanced context handling.`,match:{or:[{contains:"gemini-2.5-flash-preview-05-20"},{contains:"gemini-2.5-flash-preview-04-17"},{equals:"gemini-2.5-flash-preview-05-20:thinking"},{equals:"gemini-2.5-flash-preview"},{equals:"gemini-2.5-flash-preview:thinking"}]},price_comments:"from https://cloud.google.com/vertex-ai/generative-ai/pricing should be retired 2025-07-15",prices:{input_mtok:.15,output_mtok:.6},deprecated:!0},{id:"gemini-2.5-pro",name:"Gemini 2.5 Pro",description:`Gemini 2.5 Pro is Google's state-of-the-art AI model designed for advanced reasoning, coding, mathematics, and scientific tasks. It employs "thinking" capabilities, enabling it to reason through responses with enhanced accuracy and nuanced context handling. Gemini 2.5 Pro achieves top-tier performance on multiple benchmarks, including first-place positioning on the LMArena leaderboard, reflecting superior human-preference alignment and complex problem-solving abilities.`,match:{starts_with:"gemini-2.5-pro"},price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro",prices:{input_mtok:{base:1.25,tiers:[{start:2e5,price:2.5}]},cache_read_mtok:{base:.125,tiers:[{start:2e5,price:.25}]},output_mtok:{base:10,tiers:[{start:2e5,price:15}]}}},{id:"gemini-3-flash-preview",name:"Gemini 3 Flash Preview",description:"Google's ultra-fast frontier model optimized for speed and efficiency. Delivers state-of-the-art performance while maintaining low latency and cost, with improved reasoning and coding capabilities.",match:{or:[{equals:"gemini-3-flash-preview"},{starts_with:"gemini-3-flash-preview-"}]},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing. Standard pricing shown; Batch API offers 50% discount on input/output.",prices:{input_mtok:.5,cache_read_mtok:.05,output_mtok:3,input_audio_mtok:1,cache_audio_read_mtok:.1}},{id:"gemini-3-pro-image-preview",name:"Gemini 3 Pro Image Preview",description:"Google's image generation model optimized for high-quality image generation. Supports 1K/2K and 4K resolution outputs with flexible pricing based on image dimensions.",match:{or:[{starts_with:"gemini-3-pro-image-preview"},{equals:"gemini-3-pro-image-preview"}]},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image. Image output is priced at $120 per 1M tokens, with each 1K/2K image = 1120 tokens = $0.134/image and each 4K image = 2000 tokens = $0.24/image.",prices:{input_mtok:2,output_mtok:120}},{id:"gemini-3-pro-preview",name:"Gemini 3 Pro Preview",description:"The best model in the world for multimodal understanding, and our most powerful agentic and vibe-coding model yet.",match:{or:[{starts_with:"gemini-3-pro-preview"},{equals:"gemini-3-pro-text-preview"}]},prices:{input_mtok:{base:2,tiers:[{start:2e5,price:4}]},cache_read_mtok:{base:.2,tiers:[{start:2e5,price:.4}]},output_mtok:{base:12,tiers:[{start:2e5,price:18}]}}},{id:"gemini-3.1-flash-image-preview",name:"Gemini 3.1 Flash Image Preview",description:"Google's latest image generation model (Nano Banana 2) optimized for fast, high-quality image generation. Supports multiple output resolutions from 512px to 4K, with text and thinking output priced separately from image output tokens.",match:{starts_with:"gemini-3.1-flash-image-preview"},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing. Image output is priced at $60 per 1M tokens. Preview model - pricing may change.",prices:{input_mtok:.5,output_mtok:60}},{id:"gemini-3.1-flash-lite",name:"Gemini 3.1 Flash Lite",description:"Google's fastest and most cost-efficient Gemini 3 series model, built for intelligence at scale. Optimized for high-volume, low-latency applications while maintaining strong multimodal capabilities.",match:{starts_with:"gemini-3.1-flash-lite"},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing.",prices:{input_mtok:.25,cache_read_mtok:.025,output_mtok:1.5,input_audio_mtok:.5,cache_audio_read_mtok:.05}},{id:"gemini-3.1-pro-preview",name:"Gemini 3.1 Pro Preview",description:"The latest performance, intelligence, and usability improvements to the best model family in the world for multimodal understanding, agentic capabilities, and vibe-coding.",match:{starts_with:"gemini-3.1-pro-preview"},prices:{input_mtok:{base:2,tiers:[{start:2e5,price:4}]},cache_read_mtok:{base:.2,tiers:[{start:2e5,price:.4}]},output_mtok:{base:12,tiers:[{start:2e5,price:18}]}}},{id:"gemini-embedding-001",match:{equals:"gemini-embedding-001"},prices:{input_mtok:.15}},{id:"gemini-flash-1.5",name:"Gemini 1.5 Flash",description:"Gemini 1.5 Flash is a foundation model that performs well at a variety of multimodal tasks such as visual understanding, classification, summarization, and creating content from image, audio and video. It's adept at processing visual and text inputs such as photographs, documents, infographics, and screenshots.",match:{equals:"gemini-flash-1.5"},price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-flash",prices:{input_mtok:{base:.075,tiers:[{start:128e3,price:.15}]},cache_read_mtok:{base:.01875,tiers:[{start:128e3,price:.0375}]},output_mtok:{base:.3,tiers:[{start:128e3,price:.6}]}}},{id:"gemini-flash-1.5-8b",name:"gemini 1.5 flash",description:"A faster, more cost-effective variant of Gemini 1.5 that maintains strong capabilities while optimizing for performance and cost efficiency. Suitable for production deployments requiring high throughput.",match:{equals:"gemini-flash-1.5-8b"},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-flash-8b",prices:{input_mtok:{base:.0375,tiers:[{start:128e3,price:.075}]},cache_read_mtok:{base:.01,tiers:[{start:128e3,price:.02}]},output_mtok:{base:.15,tiers:[{start:128e3,price:.3}]}}},{id:"gemini-live-2.5-flash-preview",match:{or:[{starts_with:"gemini-live-2.5-flash-preview"},{starts_with:"gemini-2.5-flash-native-audio-preview"}]},prices:{input_mtok:.5,output_mtok:2,input_audio_mtok:3,output_audio_mtok:12}},{id:"gemini-pro",name:"gemini 1.0 pro",description:"Google's first-generation advanced multimodal model that can understand text, code, and images. It provides strong reasoning capabilities and follows instructions effectively.",match:{or:[{equals:"gemini-pro"},{equals:"gemini-1.0-pro"}]},context_window:32768,price_comments:"I can't find anything so trusting these prices, not sure the model still exists",prices:{input_mtok:.125,output_mtok:.375}},{id:"gemini-pro-1.5",name:"Gemini 1.5 Pro",description:"Google's latest multimodal model, supports image and video[0] in text or chat prompts.",match:{equals:"gemini-pro-1.5"},context_window:2e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-pro",prices:{input_mtok:{base:1.25,tiers:[{start:128e3,price:2.5}]},cache_read_mtok:{base:.3125,tiers:[{start:128e3,price:.625}]},output_mtok:{base:5,tiers:[{start:128e3,price:10}]}}},{id:"gemma-3",name:"Gemma 3 (free)",description:"Lightweight, state-of the art, open model built from the same technology that powers our Gemini models.",match:{or:[{starts_with:"gemma-3-"},{equals:"gemma-3"}]},prices:{}},{id:"gemma-3n",name:"Gemma 3n (free)",description:"Our open model built for efficient performance on everyday devices like mobile phones, laptops, and tablets.",match:{or:[{starts_with:"gemma-3n"}]},prices:{}}]},{id:"groq",name:"Groq",pricing_urls:["https://groq.com/pricing/"],api_pattern:"https://api\\.groq\\.com",extractors:[{api_flavor:"default",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"deepseek-r1-distill-llama-70b",name:"DeepSeek R1 Distill Llama 70B",match:{equals:"deepseek-r1-distill-llama-70b"},context_window:131072,prices:{input_mtok:.75,output_mtok:.99}},{id:"gemma-7b-it",match:{equals:"gemma-7b-it"},prices:{input_mtok:.07,output_mtok:.07}},{id:"gemma2-9b-it",name:"Gemma 2 9B 8k",match:{or:[{equals:"gemma2-9b-it"},{equals:"gemma2-9b"}]},prices:{input_mtok:.2,output_mtok:.2}},{id:"llama-3.1-405b-reasoning",match:{equals:"llama-3.1-405b-reasoning"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama-3.1-70b-versatile",match:{equals:"llama-3.1-70b-versatile"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama-3.1-8b-instant",name:"Llama 3.1 8B Instant 128k",match:{equals:"llama-3.1-8b-instant"},prices:{input_mtok:.05,output_mtok:.08}},{id:"llama-3.2-11b-text-preview",match:{equals:"llama-3.2-11b-text-preview"},prices:{input_mtok:.18,output_mtok:.18}},{id:"llama-3.2-11b-vision-preview",match:{equals:"llama-3.2-11b-vision-preview"},prices:{input_mtok:.18,output_mtok:.18}},{id:"llama-3.2-1b-preview",match:{equals:"llama-3.2-1b-preview"},prices:{input_mtok:.04,output_mtok:.04}},{id:"llama-3.2-3b-preview",match:{equals:"llama-3.2-3b-preview"},prices:{input_mtok:.06,output_mtok:.06}},{id:"llama-3.2-90b-text-preview",match:{equals:"llama-3.2-90b-text-preview"},prices:{input_mtok:.9,output_mtok:.9}},{id:"llama-3.2-90b-vision-preview",match:{equals:"llama-3.2-90b-vision-preview"},prices:{input_mtok:.9,output_mtok:.9}},{id:"llama-3.3-70b-specdec",match:{equals:"llama-3.3-70b-specdec"},prices:{input_mtok:.59,output_mtok:.99}},{id:"llama-3.3-70b-versatile",name:"Llama 3.3 70B Versatile 128k",match:{equals:"llama-3.3-70b-versatile"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama-guard-3-8b",match:{equals:"llama-guard-3-8b"},prices:{input_mtok:.2,output_mtok:.2}},{id:"llama2-70b-4096",match:{equals:"llama2-70b-4096"},prices:{input_mtok:.7,output_mtok:.8}},{id:"llama3-70b-8192",match:{equals:"llama3-70b-8192"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama3-8b-8192",match:{equals:"llama3-8b-8192"},prices:{input_mtok:.05,output_mtok:.08}},{id:"llama3-groq-70b-8192-tool-use-preview",match:{equals:"llama3-groq-70b-8192-tool-use-preview"},prices:{input_mtok:.89,output_mtok:.89}},{id:"llama3-groq-8b-8192-tool-use-preview",match:{equals:"llama3-groq-8b-8192-tool-use-preview"},prices:{input_mtok:.19,output_mtok:.19}},{id:"meta-llama/llama-4-maverick-17b-128e-instruct",name:"Llama 4 Maverick 17B 128E",match:{equals:"meta-llama/llama-4-maverick-17b-128e-instruct"},context_window:131072,prices:{input_mtok:.2,output_mtok:.6}},{id:"meta-llama/llama-4-scout-17b-16e-instruct",name:"Llama 4 Scout (17Bx16E) 128k",match:{equals:"meta-llama/llama-4-scout-17b-16e-instruct"},prices:{input_mtok:.11,output_mtok:.34}},{id:"meta-llama/llama-guard-4-12b",name:"Llama Guard 4 12B",match:{equals:"meta-llama/llama-guard-4-12b"},context_window:131072,prices:{input_mtok:.2,output_mtok:.2}},{id:"mistral-saba-24b",match:{equals:"mistral-saba-24b"},prices:{input_mtok:.79,output_mtok:.79}},{id:"mixtral-8x7b-32768",match:{equals:"mixtral-8x7b-32768"},prices:{input_mtok:.24,output_mtok:.24}},{id:"moonshotai/kimi-k2-instruct",name:"Kimi K2 1T 128k",match:{or:[{equals:"moonshotai/kimi-k2-instruct"},{equals:"moonshotai/kimi-k2-instruct-0905"}]},context_window:131072,prices:{input_mtok:1,cache_read_mtok:.5,output_mtok:3}},{id:"openai/gpt-oss-120b",description:`GPT-OSS 120B is OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with
+`,match:{or:[{starts_with:"claude-2"},{contains:"claude-v2"}]},context_window:2e5,prices:{input_mtok:8,output_mtok:24}},{id:"claude-3-5-haiku-latest",name:"Claude Haiku 3.5",description:"Fastest, most cost-effective model",match:{or:[{starts_with:"claude-3-5-haiku"},{starts_with:"claude-3.5-haiku"}]},context_window:2e5,prices:{input_mtok:.8,cache_write_mtok:1,cache_read_mtok:.08,output_mtok:4}},{id:"claude-3-5-sonnet",name:"Claude Sonnet 3.5",description:"Claude 3.5 Sonnet is an ideal balance of intelligence and speed for enterprise workloads. Maximum utility at a lower price, dependable, balanced for scaled deployments.",match:{or:[{starts_with:"claude-3-5-sonnet"},{starts_with:"claude-3.5-sonnet"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-7-sonnet-latest",name:"Claude Sonnet 3.7",description:"Claude 3.7 Sonnet is an advanced large language model with improved reasoning, coding, and problem-solving capabilities.",match:{or:[{starts_with:"claude-3-7-sonnet"},{starts_with:"claude-3.7-sonnet"},{starts_with:"claude-sonnet-3.7"},{starts_with:"claude-sonnet-3-7"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-haiku",name:"Claude Haiku 3",description:"Fastest, most cost-effective model",match:{starts_with:"claude-3-haiku"},context_window:2e5,prices:{input_mtok:.25,cache_write_mtok:.3,cache_read_mtok:.03,output_mtok:1.25}},{id:"claude-3-opus-latest",name:"Claude Opus 3",description:"Claude 3 Opus was Anthropic's most powerful model for highly complex tasks. It boasts top-level performance, intelligence, fluency, and understanding.",match:{starts_with:"claude-3-opus"},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-3-sonnet",name:"Claude 3 Sonnet",description:"Claude 3 Sonnet is an ideal balance of intelligence and speed for enterprise workloads. Maximum utility at a lower price, dependable, balanced for scaled deployments.",match:{starts_with:"claude-3-sonnet"},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-haiku-4-5",name:"Claude Haiku 4.5",description:"Fastest and most intelligent Haiku model",match:{or:[{starts_with:"claude-haiku-4-5"},{starts_with:"claude-haiku-4.5"},{starts_with:"claude-4-5-haiku"},{starts_with:"claude-4.5-haiku"}]},context_window:2e5,prices:{input_mtok:1,cache_write_mtok:1.25,cache_read_mtok:.1,output_mtok:5}},{id:"claude-opus-4-0",name:"Claude Opus 4",description:"Most intelligent model for complex tasks",match:{or:[{starts_with:"claude-opus-4-0"},{starts_with:"claude-4-opus"},{equals:"claude-opus-4"},{equals:"claude-opus-4-20250514"}]},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-opus-4-1",name:"Claude Opus 4.1",description:"Most intelligent model for complex tasks",match:{or:[{starts_with:"claude-opus-4-1"},{starts_with:"claude-opus-4.1"}]},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-opus-4-5",name:"Claude Opus 4.5",description:"Premium model combining maximum intelligence with practical performance",match:{or:[{starts_with:"claude-opus-4-5"},{starts_with:"claude-opus-4.5"},{starts_with:"claude-4-5-opus"},{starts_with:"claude-4.5-opus"}]},context_window:2e5,prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"claude-opus-4-6",name:"Claude Opus 4.6",description:"Our most intelligent model for building agents and coding",match:{or:[{starts_with:"claude-opus-4-6"},{starts_with:"claude-opus-4.6"},{starts_with:"claude-4-6-opus"},{starts_with:"claude-4.6-opus"}]},context_window:2e5,prices:[{prices:{input_mtok:{base:5,tiers:[{start:2e5,price:10}]},cache_write_mtok:{base:6.25,tiers:[{start:2e5,price:12.5}]},cache_read_mtok:{base:.5,tiers:[{start:2e5,price:1}]},output_mtok:{base:25,tiers:[{start:2e5,price:37.5}]}}},{constraint:{start_date:"2026-03-13",type:"start_date"},prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}}]},{id:"claude-opus-4-7",name:"Claude Opus 4.7",description:"Our most capable model for complex reasoning and agentic coding",match:{or:[{starts_with:"claude-opus-4-7"},{starts_with:"claude-opus-4.7"},{starts_with:"claude-4-7-opus"},{starts_with:"claude-4.7-opus"}]},context_window:1e6,price_comments:"Flat pricing across full 1M context window (no tiered pricing). Ref: https://platform.claude.com/docs/en/about-claude/pricing#long-context-pricing",prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"claude-sonnet-4-0",name:"Claude Sonnet 4",description:"Optimal balance of intelligence, cost, and speed",match:{or:[{starts_with:"claude-sonnet-4-2025"},{starts_with:"claude-sonnet-4-0"},{starts_with:"claude-sonnet-4@"},{equals:"claude-sonnet-4"},{starts_with:"claude-4-sonnet"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-sonnet-4-5",name:"Claude Sonnet 4.5",description:"Our best combination of speed and intelligence",match:{or:[{starts_with:"claude-sonnet-4-5"},{starts_with:"claude-sonnet-4.5"}]},context_window:1e6,prices:{input_mtok:{base:3,tiers:[{start:2e5,price:6}]},cache_write_mtok:{base:3.75,tiers:[{start:2e5,price:7.5}]},cache_read_mtok:{base:.3,tiers:[{start:2e5,price:.6}]},output_mtok:{base:15,tiers:[{start:2e5,price:22.5}]}}},{id:"claude-sonnet-4-6",name:"Claude Sonnet 4.6",description:"Our best combination of speed and intelligence",match:{or:[{starts_with:"claude-sonnet-4-6"},{starts_with:"claude-sonnet-4.6"}]},context_window:1e6,prices:[{prices:{input_mtok:{base:3,tiers:[{start:2e5,price:6}]},cache_write_mtok:{base:3.75,tiers:[{start:2e5,price:7.5}]},cache_read_mtok:{base:.3,tiers:[{start:2e5,price:.6}]},output_mtok:{base:15,tiers:[{start:2e5,price:22.5}]}}},{constraint:{start_date:"2026-03-13",type:"start_date"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}}]},{id:"claude-v1",description:"Retired, here to match price sources",match:{equals:"claude-v1"},prices:{input_mtok:8,output_mtok:24}}]},{id:"avian",name:"Avian",pricing_urls:["https://avian.io/pricing/"],api_pattern:"https://api\\.avian\\.io",models:[{id:"Meta-Llama-3.1-405B-Instruct",match:{equals:"Meta-Llama-3.1-405B-Instruct"},prices:{input_mtok:1.5,output_mtok:1.5}},{id:"Meta-Llama-3.1-70B-Instruct",match:{equals:"Meta-Llama-3.1-70B-Instruct"},prices:{input_mtok:.45,output_mtok:.45}},{id:"Meta-Llama-3.1-8B-Instruct",match:{equals:"Meta-Llama-3.1-8B-Instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"Meta-Llama-3.3-70B-Instruct",match:{equals:"Meta-Llama-3.3-70B-Instruct"},prices:{input_mtok:.45,output_mtok:.45}}]},{id:"aws",name:"AWS Bedrock",pricing_urls:["https://aws.amazon.com/bedrock/pricing/"],api_pattern:"https://bedrock-runtime\\.[a-z0-9-]+\\.amazonaws\\.com/",provider_match:{or:[{contains:"bedrock"},{contains:"amazon"}]},extractors:[{api_flavor:"default",root:"usage",model_path:"model",mappings:[{path:"inputTokens",dest:"input_tokens",required:!0},{path:"outputTokens",dest:"output_tokens",required:!0}]},{api_flavor:"anthropic",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"cache_creation_input_tokens",dest:"input_tokens",required:!1},{path:"cache_read_input_tokens",dest:"input_tokens",required:!1},{path:"cache_creation_input_tokens",dest:"cache_write_tokens",required:!1},{path:"cache_read_input_tokens",dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"amazon.nova-lite-v1:0",name:"Nova Lite",description:"Amazon Nova Lite 1.0 is a very low-cost multimodal model from Amazon that focused on fast processing of image, video, and text inputs to generate text output. Amazon Nova Lite can handle real-time customer interactions, document analysis, and visual question-answering tasks with high accuracy.",match:{contains:"amazon.nova-lite"},prices:{input_mtok:.06,cache_read_mtok:.015,output_mtok:.24}},{id:"amazon.nova-micro-v1:0",name:"Nova Micro",description:"Amazon Nova Micro 1.0 is a text-only model that delivers the lowest latency responses in the Amazon Nova family of models at a very low cost. With a context length of 128K tokens and optimized for speed and cost, Amazon Nova Micro excels at tasks such as text summarization, translation, content classification, interactive chat, and brainstorming. It has  simple mathematical reasoning and coding abilities.",match:{contains:"amazon.nova-micro"},prices:{input_mtok:.035,cache_read_mtok:.00875,output_mtok:.14}},{id:"amazon.nova-premier-v1:0",name:"Nova Premier",match:{contains:"amazon.nova-premier"},prices:{input_mtok:2.5,cache_read_mtok:.625,output_mtok:12.5}},{id:"amazon.nova-pro-v1:0",name:"Nova Pro",description:"Amazon Nova Pro 1.0 is a capable multimodal model from Amazon focused on providing a combination of accuracy, speed, and cost for a wide range of tasks. As of December 2024, it achieves state-of-the-art performance on key benchmarks including visual question answering (TextVQA) and video understanding (VATEX).",match:{contains:"amazon.nova-pro"},prices:{input_mtok:.8,cache_read_mtok:.2,output_mtok:3.2}},{id:"amazon.nova-sonic-v1:0",name:"Nova Sonic",match:{contains:"amazon.nova-sonic"},prices:{input_mtok:.06,output_mtok:.24,input_audio_mtok:3.4,output_audio_mtok:13.6}},{id:"amazon.titan-embed-text-v1",name:"Titan Embeddings G1 - Text",match:{contains:"amazon.titan-embed-text"},prices:{input_mtok:.1}},{id:"amazon.titan-text-express-v1",name:"Titan Text G1 - Express",match:{contains:"titan-text-express"},prices:{input_mtok:.2,output_mtok:.6}},{id:"amazon.titan-text-lite-v1",name:"Titan Text G1 - Lite",match:{contains:"titan-text-lite"},prices:{input_mtok:.15,output_mtok:.2}},{id:"deepseek.r1-v1:0",name:"DeepSeek-R1",match:{contains:"deepseek.r1"},prices:{input_mtok:1.35,output_mtok:5.4}},{id:"global.anthropic.claude-haiku-4-5-20251001-v1:0",match:{contains:"global.anthropic.claude-haiku-4-5-20251001"},prices:{input_mtok:1,cache_write_mtok:1.25,cache_read_mtok:.1,output_mtok:5}},{id:"global.anthropic.claude-opus-4-5-v1:0",match:{contains:"global.anthropic.claude-opus-4-5"},prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"global.anthropic.claude-opus-4-6-v1:0",match:{contains:"global.anthropic.claude-opus-4-6"},prices:{input_mtok:{base:5,tiers:[{start:2e5,price:10}]},cache_write_mtok:{base:6.25,tiers:[{start:2e5,price:12.5}]},cache_read_mtok:{base:.5,tiers:[{start:2e5,price:1}]},output_mtok:{base:25,tiers:[{start:2e5,price:37.5}]}}},{id:"global.anthropic.claude-opus-4-7-v1:0",match:{contains:"global.anthropic.claude-opus-4-7"},prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"global.anthropic.claude-sonnet-4-20250514-v1:0",match:{contains:"global.anthropic.claude-sonnet-4-20250514"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"global.anthropic.claude-sonnet-4-5-20250929-v1:0",match:{contains:"global.anthropic.claude-sonnet-4-5-20250929"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"global.anthropic.claude-sonnet-4-6-v1:0",match:{contains:"global.anthropic.claude-sonnet-4-6"},prices:{input_mtok:{base:3,tiers:[{start:2e5,price:6}]},cache_write_mtok:{base:3.75,tiers:[{start:2e5,price:7.5}]},cache_read_mtok:{base:.3,tiers:[{start:2e5,price:.6}]},output_mtok:{base:15,tiers:[{start:2e5,price:22.5}]}}},{id:"meta.llama3-1-70b-instruct-v1:0",name:"Llama 3.1 70B Instruct",match:{contains:"meta.llama3-1-70b-instruct"},prices:{input_mtok:.72,output_mtok:.72}},{id:"meta.llama3-1-8b-instruct-v1:0",name:"Llama 3.1 8B Instruct",match:{contains:"meta.llama3-1-8b-instruct"},prices:{input_mtok:.22,output_mtok:.22}},{id:"meta.llama3-2-11b-instruct-v1:0",name:"Llama 3.2 11B Instruct",match:{contains:"meta.llama3-2-11b-instruct"},prices:{input_mtok:.16,output_mtok:.16}},{id:"meta.llama3-2-1b-instruct-v1:0",name:"Llama 3.2 1B Instruct",match:{contains:"meta.llama3-2-1b-instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"meta.llama3-2-3b-instruct-v1:0",name:"Llama 3.2 3B Instruct",match:{contains:"meta.llama3-2-3b-instruct"},prices:{input_mtok:.15,output_mtok:.15}},{id:"meta.llama3-2-90b-instruct-v1:0",name:"Llama 3.2 90B Instruct",match:{contains:"meta.llama3-2-90b-instruct"},prices:{input_mtok:.72,output_mtok:.72}},{id:"meta.llama3-3-70b-instruct-v1:0",name:"Llama 3.3 70B Instruct",match:{contains:"meta.llama3-3-70b-instruct"},prices:{input_mtok:.72,output_mtok:.72}},{id:"meta.llama3-70b-instruct-v1:0",name:"Llama 3 70B Instruct",match:{contains:"meta.llama3-70b-instruct"},prices:{input_mtok:2.65,output_mtok:3.5}},{id:"meta.llama3-8b-instruct-v1:0",name:"Llama 3 8B Instruct",match:{contains:"meta.llama3-8b-instruct"},prices:{input_mtok:.3,output_mtok:.6}},{id:"meta.llama4-maverick-17b-instruct-v1:0",name:"Llama 4 Maverick 17B Instruct",match:{contains:"meta.llama4-maverick-17b-instruct"},prices:{input_mtok:.24,output_mtok:.97}},{id:"meta.llama4-scout-17b-instruct-v1:0",name:"Llama 4 Scout 17B Instruct",match:{contains:"meta.llama4-scout-17b-instruct"},prices:{input_mtok:.17,output_mtok:.66}},{id:"mistral.mistral-7b-instruct-v0:2",name:"Mistral 7B Instruct",match:{contains:"mistral.mistral-7b-instruct-v0"},prices:{input_mtok:.15,output_mtok:.2}},{id:"mistral.mistral-large-2402-v1:0",name:"Mistral Large (24.02)",match:{contains:"mistral.mistral-large-2402"},prices:{input_mtok:4,output_mtok:12}},{id:"mistral.mistral-small-2402-v1:0",name:"Mistral Small (24.02)",match:{contains:"mistral.mistral-small-2402"},prices:{input_mtok:1,output_mtok:3}},{id:"mistral.mixtral-8x7b-instruct-v0:1",name:"Mixtral 8x7B Instruct",match:{contains:"mistral.mixtral-8x7b-instruct-v0"},prices:{input_mtok:.45,output_mtok:.7}},{id:"mistral.pixtral-large-2502-v1:0",name:"Pixtral Large (25.02)",match:{contains:"mistral.pixtral-large-2502"},prices:{input_mtok:2,output_mtok:6}},{id:"nvidia.nemotron-nano-3-30b:0",name:"Nemotron 3 Nano 30B",match:{contains:"nvidia.nemotron-nano-3-30b"},prices:{input_mtok:.06,output_mtok:.24}},{id:"nvidia.nemotron-nano-9b-v2:0",name:"Nemotron 2 Nano 9B",match:{contains:"nvidia.nemotron-nano-9b-v2"},prices:{input_mtok:.06,output_mtok:.23}},{id:"nvidia.nemotron-super-3-120b:0",name:"Nemotron 3 Super 120B",match:{contains:"nvidia.nemotron-super-3-120b"},prices:{input_mtok:.15,output_mtok:.65}},{id:"openai.gpt-oss-120b-1:0",name:"gpt-oss-120b",match:{contains:"openai.gpt-oss-120b-1"},prices:{input_mtok:.15,output_mtok:.6}},{id:"openai.gpt-oss-20b-1:0",name:"gpt-oss-20b",match:{contains:"openai.gpt-oss-20b-1"},prices:{input_mtok:.07,output_mtok:.3}},{id:"qwen.qwen3-32b-v1:0",name:"Qwen3 32B (dense)",match:{contains:"qwen.qwen3-32b"},prices:{input_mtok:.15,output_mtok:.6}},{id:"qwen.qwen3-coder-30b-a3b-v1:0",name:"Qwen3-Coder-30B-A3B-Instruct",match:{contains:"qwen.qwen3-coder-30b-a3b"},prices:{input_mtok:.15,output_mtok:.6}},{id:"qwen.qwen3-coder-480b-a35b-v1:0",name:"Qwen3-Coder-480B-A35B-Instruct",match:{contains:"qwen.qwen3-coder-480b-a35b"},prices:{input_mtok:.45,output_mtok:1.8}},{id:"qwen.qwen3-vl-235b-a22b-v1:0",name:"Qwen3-VL-235B-A22B-Instruct",match:{contains:"qwen.qwen3-vl-235b-a22b"},prices:{input_mtok:.53,output_mtok:2.66}},{id:"regional.anthropic.claude-3-5-haiku-20241022-v1:0",match:{contains:"claude-3-5-haiku-20241022"},prices:{input_mtok:.8,cache_write_mtok:1,cache_read_mtok:.08,output_mtok:4}},{id:"regional.anthropic.claude-3-5-sonnet-20240620-v1:0",match:{contains:"claude-3-5-sonnet-20240620"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-3-5-sonnet-20241022-v2:0",match:{contains:"claude-3-5-sonnet-20241022"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-3-7-sonnet-20250219-v1:0",match:{contains:"claude-3-7-sonnet-20250219"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-3-haiku-20240307-v1:0",match:{contains:"claude-3-haiku-20240307"},prices:{input_mtok:.25,output_mtok:1.25}},{id:"regional.anthropic.claude-3-opus-20240229-v1:0",match:{contains:"claude-3-opus-20240229"},prices:{input_mtok:15,output_mtok:75}},{id:"regional.anthropic.claude-3-sonnet-20240229-v1:0",match:{contains:"claude-3-sonnet-20240229"},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-haiku-4-5-20251001-v1:0",match:{or:[{starts_with:"anthropic.claude-haiku-4-5-20251001"},{starts_with:"claude-haiku-4-5-20251001"},{contains:"us.anthropic.claude-haiku-4-5-20251001"},{contains:"au.anthropic.claude-haiku-4-5-20251001"},{contains:"apac.anthropic.claude-haiku-4-5-20251001"},{contains:"eu.anthropic.claude-haiku-4-5-20251001"},{contains:"us-gov.anthropic.claude-haiku-4-5-20251001"},{contains:"jp.anthropic.claude-haiku-4-5-20251001"}]},prices:{input_mtok:1.1,cache_write_mtok:1.375,cache_read_mtok:.11,output_mtok:5.5}},{id:"regional.anthropic.claude-opus-4-1-20250805-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-1-20250805"},{starts_with:"claude-opus-4-1-20250805"},{contains:"us.anthropic.claude-opus-4-1-20250805"},{contains:"au.anthropic.claude-opus-4-1-20250805"},{contains:"apac.anthropic.claude-opus-4-1-20250805"},{contains:"eu.anthropic.claude-opus-4-1-20250805"},{contains:"us-gov.anthropic.claude-opus-4-1-20250805"},{contains:"jp.anthropic.claude-opus-4-1-20250805"}]},prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"regional.anthropic.claude-opus-4-20250514-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-20250514"},{starts_with:"claude-opus-4-20250514"},{contains:"us.anthropic.claude-opus-4-20250514"},{contains:"au.anthropic.claude-opus-4-20250514"},{contains:"apac.anthropic.claude-opus-4-20250514"},{contains:"eu.anthropic.claude-opus-4-20250514"},{contains:"us-gov.anthropic.claude-opus-4-20250514"},{contains:"jp.anthropic.claude-opus-4-20250514"}]},prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"regional.anthropic.claude-opus-4-5-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-5"},{starts_with:"claude-opus-4-5"},{contains:"us.anthropic.claude-opus-4-5"},{contains:"au.anthropic.claude-opus-4-5"},{contains:"apac.anthropic.claude-opus-4-5"},{contains:"eu.anthropic.claude-opus-4-5"},{contains:"us-gov.anthropic.claude-opus-4-5"},{contains:"jp.anthropic.claude-opus-4-5"}]},prices:{input_mtok:5.5,cache_write_mtok:6.875,cache_read_mtok:.55,output_mtok:27.5}},{id:"regional.anthropic.claude-opus-4-6-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-6"},{starts_with:"claude-opus-4-6"},{contains:"us.anthropic.claude-opus-4-6"},{contains:"au.anthropic.claude-opus-4-6"},{contains:"apac.anthropic.claude-opus-4-6"},{contains:"eu.anthropic.claude-opus-4-6"},{contains:"us-gov.anthropic.claude-opus-4-6"},{contains:"jp.anthropic.claude-opus-4-6"}]},prices:{input_mtok:{base:5.5,tiers:[{start:2e5,price:11}]},cache_write_mtok:{base:6.875,tiers:[{start:2e5,price:13.75}]},cache_read_mtok:{base:.55,tiers:[{start:2e5,price:1.1}]},output_mtok:{base:27.5,tiers:[{start:2e5,price:41.25}]}}},{id:"regional.anthropic.claude-opus-4-7-v1:0",match:{or:[{starts_with:"anthropic.claude-opus-4-7"},{starts_with:"claude-opus-4-7"},{contains:"us.anthropic.claude-opus-4-7"},{contains:"au.anthropic.claude-opus-4-7"},{contains:"apac.anthropic.claude-opus-4-7"},{contains:"eu.anthropic.claude-opus-4-7"},{contains:"us-gov.anthropic.claude-opus-4-7"},{contains:"jp.anthropic.claude-opus-4-7"}]},prices:{input_mtok:5.5,cache_write_mtok:6.875,cache_read_mtok:.55,output_mtok:27.5}},{id:"regional.anthropic.claude-sonnet-4-20250514-v1:0",match:{or:[{starts_with:"anthropic.claude-sonnet-4-20250514"},{starts_with:"claude-sonnet-4-20250514"},{contains:"us.anthropic.claude-sonnet-4-20250514"},{contains:"au.anthropic.claude-sonnet-4-20250514"},{contains:"apac.anthropic.claude-sonnet-4-20250514"},{contains:"eu.anthropic.claude-sonnet-4-20250514"},{contains:"us-gov.anthropic.claude-sonnet-4-20250514"},{contains:"jp.anthropic.claude-sonnet-4-20250514"}]},prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"regional.anthropic.claude-sonnet-4-5-20250929-v1:0",match:{or:[{starts_with:"anthropic.claude-sonnet-4-5-20250929"},{starts_with:"claude-sonnet-4-5-20250929"},{contains:"us.anthropic.claude-sonnet-4-5-20250929"},{contains:"au.anthropic.claude-sonnet-4-5-20250929"},{contains:"apac.anthropic.claude-sonnet-4-5-20250929"},{contains:"eu.anthropic.claude-sonnet-4-5-20250929"},{contains:"us-gov.anthropic.claude-sonnet-4-5-20250929"},{contains:"jp.anthropic.claude-sonnet-4-5-20250929"}]},prices:{input_mtok:3.3,cache_write_mtok:4.125,cache_read_mtok:.33,output_mtok:16.5}},{id:"regional.anthropic.claude-sonnet-4-6-v1:0",match:{or:[{starts_with:"anthropic.claude-sonnet-4-6"},{starts_with:"claude-sonnet-4-6"},{contains:"us.anthropic.claude-sonnet-4-6"},{contains:"au.anthropic.claude-sonnet-4-6"},{contains:"apac.anthropic.claude-sonnet-4-6"},{contains:"eu.anthropic.claude-sonnet-4-6"},{contains:"us-gov.anthropic.claude-sonnet-4-6"},{contains:"jp.anthropic.claude-sonnet-4-6"}]},prices:{input_mtok:{base:3.3,tiers:[{start:2e5,price:6.6}]},cache_write_mtok:{base:4.125,tiers:[{start:2e5,price:8.25}]},cache_read_mtok:{base:.33,tiers:[{start:2e5,price:.66}]},output_mtok:{base:16.5,tiers:[{start:2e5,price:24.75}]}}}]},{id:"azure",name:"Microsoft Azure",pricing_urls:["https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/#pricing"],api_pattern:"(https?://)?([^.]*\\.)?(?:openai\\.azure\\.com|azure-api\\.net|cognitiveservices\\.azure\\.com)",price_comments:'These are prices for "*-Global" models, prices for "Regional" models are often slightly higher. Retired models are listed at https://learn.microsoft.com/th-th/azure/ai-foundry/openai/concepts/legacy-models',extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["prompt_tokens_details","audio_tokens"],dest:"input_audio_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"responses",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:["input_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"embeddings",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0}]},{api_flavor:"anthropic",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"cache_creation_input_tokens",dest:"input_tokens",required:!1},{path:"cache_read_input_tokens",dest:"input_tokens",required:!1},{path:"cache_creation_input_tokens",dest:"cache_write_tokens",required:!1},{path:"cache_read_input_tokens",dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]}],fallback_model_providers:["openai","anthropic"],models:[{id:"ada",match:{or:[{equals:"ada"},{equals:"text-embedding-ada"},{equals:"text-embedding-ada-002"},{equals:"text-embedding-ada-002-v2"}]},prices:{input_mtok:.1}},{id:"babbage",match:{or:[{equals:"babbage"},{equals:"babbage-002"}]},prices:{input_mtok:.4}},{id:"curie",match:{or:[{equals:"curie"},{equals:"text-curie"},{equals:"text-curie-001"}]},prices:{input_mtok:2}},{id:"davinci",match:{or:[{equals:"davinci"},{equals:"davinci-002"},{equals:"text-davinci"},{equals:"text-davinci-002"}]},prices:{input_mtok:2}},{id:"mai-ds-r1:free",name:"MAI DS R1 (free)",description:"MAI-DS-R1 is a post-trained variant of DeepSeek-R1 developed by the Microsoft AI team to improve the model's responsiveness on previously blocked topics while enhancing its safety profile. Built on top of DeepSeek-R1's reasoning foundation, it integrates 110k examples from the Tulu-3 SFT dataset and 350k internally curated multilingual safety-alignment samples. The model retains strong reasoning, coding, and problem-solving capabilities, while unblocking a wide range of prompts previously restricted in R1.",match:{equals:"mai-ds-r1:free"},prices:{}},{id:"o1",match:{or:[{equals:"o1"},{equals:"o1-2024-12-17"},{equals:"o1-preview"},{equals:"o1-preview-2024-09-12"}]},prices:{input_mtok:15,cache_read_mtok:7.5,output_mtok:60}},{id:"o1-mini",match:{or:[{equals:"o1-mini"},{equals:"o1-mini-2024-09-12"}]},prices:{input_mtok:1.1,cache_read_mtok:.55,output_mtok:4.4}},{id:"o3-2025-04-16",match:{or:[{equals:"o3"},{equals:"o3-2025-04-16"}]},prices:{input_mtok:2,cache_read_mtok:.5,output_mtok:8}},{id:"o3-mini",match:{or:[{equals:"o3-mini"},{equals:"o3-mini-2025-01-31"}]},prices:{input_mtok:1.1,cache_read_mtok:.55,output_mtok:4.4}},{id:"o4-mini",match:{or:[{contains:"o4-mini"},{contains:"o4-mini-2025-04-16"}]},prices:{input_mtok:1.1,cache_read_mtok:.28,output_mtok:4.4}},{id:"phi-3-medium-128k-instruct",name:"Phi-3 Medium 128K Instruct",description:"Phi-3 128K Medium is a powerful 14-billion parameter model designed for advanced language understanding, reasoning, and instruction following. Optimized through supervised fine-tuning and preference adjustments, it excels in tasks involving common sense, mathematics, logical reasoning, and code processing.",match:{equals:"phi-3-medium-128k-instruct"},prices:{input_mtok:1,output_mtok:1}},{id:"phi-3-mini-128k-instruct",name:"Phi-3 Mini 128K Instruct",description:"Phi-3 Mini is a powerful 3.8B parameter model designed for advanced language understanding, reasoning, and instruction following. Optimized through supervised fine-tuning and preference adjustments, it excels in tasks involving common sense, mathematics, logical reasoning, and code processing.",match:{equals:"phi-3-mini-128k-instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"phi-3.5-mini-128k-instruct",name:"Phi-3.5 Mini 128K Instruct",description:"Phi-3.5 models are lightweight, state-of-the-art open models. These models were trained with Phi-3 datasets that include both synthetic data and the filtered, publicly available websites data, with a focus on high quality and reasoning-dense properties. Phi-3.5 Mini uses 3.8B parameters, and is a dense decoder-only transformer model using the same tokenizer as Phi-3 Mini.",match:{equals:"phi-3.5-mini-128k-instruct"},prices:{input_mtok:.1,output_mtok:.1}},{id:"phi-4",name:"Phi 4",description:"Microsoft Research Phi-4 is designed to perform well in complex reasoning tasks and can operate efficiently in situations with limited memory or where quick responses are needed.",match:{equals:"phi-4"},prices:{input_mtok:.07,output_mtok:.14}},{id:"phi-4-multimodal-instruct",name:"Phi 4 Multimodal Instruct",description:"Phi-4 Multimodal Instruct is a versatile 5.6B parameter foundation model that combines advanced reasoning and instruction-following capabilities across both text and visual inputs, providing accurate text outputs. The unified architecture enables efficient, low-latency inference, suitable for edge and mobile deployments. Phi-4 Multimodal Instruct supports text inputs in multiple languages including Arabic, Chinese, English, French, German, Japanese, Spanish, and more, with visual input optimized primarily for English. It delivers impressive performance on multimodal tasks involving mathematical, scientific, and document reasoning, providing developers and enterprises a powerful yet compact model for sophisticated interactive applications. For more information, see the Phi-4 Multimodal blog post.",match:{equals:"phi-4-multimodal-instruct"},prices:{input_mtok:.05,output_mtok:.1}},{id:"phi-4-reasoning-plus",name:"Phi 4 Reasoning Plus",description:"Phi-4-reasoning-plus is an enhanced 14B parameter model from Microsoft, fine-tuned from Phi-4 with additional reinforcement learning to boost accuracy on math, science, and code reasoning tasks. It uses the same dense decoder-only transformer architecture as Phi-4, but generates longer, more comprehensive outputs structured into a step-by-step reasoning trace and final answer.",match:{equals:"phi-4-reasoning-plus"},prices:{input_mtok:.07,output_mtok:.35}},{id:"phi-4-reasoning-plus:free",name:"Phi 4 Reasoning Plus (free)",description:"Phi-4-reasoning-plus is an enhanced 14B parameter model from Microsoft, fine-tuned from Phi-4 with additional reinforcement learning to boost accuracy on math, science, and code reasoning tasks. It uses the same dense decoder-only transformer architecture as Phi-4, but generates longer, more comprehensive outputs structured into a step-by-step reasoning trace and final answer.",match:{equals:"phi-4-reasoning-plus:free"},prices:{}},{id:"phi-4-reasoning:free",name:"Phi 4 Reasoning (free)",description:"Phi-4-reasoning is a 14B parameter dense decoder-only transformer developed by Microsoft, fine-tuned from Phi-4 to enhance complex reasoning capabilities. It uses a combination of supervised fine-tuning on chain-of-thought traces and reinforcement learning, targeting math, science, and code reasoning tasks. With a 32k context window and high inference efficiency, it is optimized for structured responses in a two-part format: reasoning trace followed by a final solution.",match:{equals:"phi-4-reasoning:free"},prices:{}},{id:"text-embedding-3-large",match:{equals:"text-embedding-3-large"},prices:{input_mtok:.13}},{id:"text-embedding-3-small",match:{equals:"text-embedding-3-small"},prices:{input_mtok:.02}},{id:"wizardlm-2-8x22b",name:"WizardLM-2 8x22B",description:"WizardLM-2 8x22B is Microsoft AI's most advanced Wizard model. It demonstrates highly competitive performance compared to leading proprietary models, and it consistently outperforms all existing state-of-the-art opensource models.",match:{equals:"wizardlm-2-8x22b"},prices:{input_mtok:.48,output_mtok:.48}}]},{id:"cerebras",name:"Cerebras",pricing_urls:["https://www.cerebras.ai/pricing#pricing","https://inference-docs.cerebras.ai/models/openai-oss"],api_pattern:"https://api\\.cerebras\\.ai",model_match:{contains:"cerebras"},provider_match:{contains:"cerebras"},extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"gpt-oss-120b",name:"GPT-OSS 120B",description:"OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with 120 billion parameters and 128 experts. Delivers frontier reasoning capabilities with record-breaking inference speeds on Cerebras hardware (~3,000 tokens/second).",match:{or:[{equals:"gpt-oss-120b"},{starts_with:"cerebras/gpt-oss-120b"},{starts_with:"cerebras:gpt-oss-120b"}]},context_window:131072,price_comments:"Developer tier pricing. Free tier: 65k context, Paid tier: 131k context.",prices:{input_mtok:.35,output_mtok:.75}},{id:"llama-3.3-70b",name:"Llama 3.3 70B",description:"Meta's enhanced 70B model delivering 405B-level accuracy. Optimized for chat, coding, instruction following, mathematics, and reasoning with high-speed inference on Cerebras hardware (~2,100 tokens/second).",match:{or:[{equals:"llama-3.3-70b"},{starts_with:"cerebras/llama-3.3-70b"},{starts_with:"cerebras:llama-3.3-70b"}]},context_window:128e3,price_comments:"Developer tier pricing. Free tier: 65k context, Paid tier: 128k context.",prices:{input_mtok:.85,output_mtok:1.2}},{id:"llama3.1-8b",name:"Llama 3.1 8B",description:"Meta's Llama 3.1 8B model for general-purpose tasks including chat, coding, and instruction following. Optimized for fast inference on Cerebras hardware (~2,200 tokens/second).",match:{or:[{equals:"llama3.1-8b"},{starts_with:"cerebras/llama3.1-8b"},{starts_with:"cerebras:llama3.1-8b"}]},context_window:32768,price_comments:"Developer tier pricing. Free tier: 8k context, Paid tier: 32k context.",prices:{input_mtok:.1,output_mtok:.1}},{id:"qwen-3-32b",name:"Qwen 3 32B",description:"Qwen's 32B parameter model with enhanced reasoning and coding capabilities. Supports both standard and reasoning modes for complex tasks, with fast inference speeds on Cerebras hardware (~2,600 tokens/second).",match:{or:[{equals:"qwen-3-32b"},{starts_with:"cerebras/qwen-3-32b"},{starts_with:"cerebras:qwen-3-32b"}]},context_window:131072,price_comments:"Developer tier pricing. Free tier: 65k context, Paid tier: 131k context.",prices:{input_mtok:.4,output_mtok:.8}},{id:"qwen-3-coder-480b",name:"qwen-3-coder-480b",match:{equals:"qwen-3-coder-480b"},price_comments:"Seems to be no longer available on cerebras, here to help with tests",prices:{}}]},{id:"cohere",name:"Cohere",pricing_urls:["https://cohere.com/pricing"],api_pattern:"https://api\\.cohere\\.ai",model_match:{starts_with:"command-"},provider_match:{contains:"cohere"},extractors:[{api_flavor:"default",root:["usage","billed_units"],model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"output_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"embeddings",root:["meta","billed_units"],model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0}]}],models:[{id:"command",name:"Command",description:"Command is an instruction-following conversational model that performs language tasks with high quality, more reliably and with a longer context than our base generative models.",match:{equals:"command"},prices:{input_mtok:1,output_mtok:2}},{id:"command-a",name:"Command A",description:`Command A is an open-weights 111B parameter model with a 256k context window focused on delivering great performance across agentic, multilingual, and coding use cases.
+Compared to other leading proprietary and open-weights models Command A delivers maximum performance with minimum hardware costs, excelling on business-critical agentic and multilingual tasks.`,match:{starts_with:"command-a"},prices:{input_mtok:2.5,output_mtok:10}},{id:"command-r",name:"Command R",description:"Command-R is a 35B parameter model that performs conversational language tasks at a higher quality, more reliably, and with a longer context than previous models. It can be used for complex workflows like code generation, retrieval augmented generation (RAG), tool use, and agents.",match:{or:[{equals:"command-r"},{equals:"command-r-08-2024"}]},prices:{input_mtok:.15,output_mtok:.6}},{id:"command-r-plus",name:"Command R+",description:"Command R+ is a new, 104B-parameter LLM from Cohere. It's useful for roleplay, general consumer usecases, and Retrieval Augmented Generation (RAG).",match:{or:[{equals:"command-r-plus"},{equals:"command-r-plus-08-2024"}]},prices:{input_mtok:2.5,output_mtok:10}},{id:"command-r7b",name:"Command R7B",description:"Command R7B (12-2024) is a small, fast update of the Command R+ model, delivered in December 2024. It excels at RAG, tool use, agents, and similar tasks requiring complex reasoning and multiple steps.",match:{or:[{equals:"command-r7b"},{equals:"command-r7b-12-2024"}]},prices:{input_mtok:.0375,output_mtok:.15}},{id:"embed-v4.0",name:"Embed v4.0",description:"Embed v4.0 is a state-of-the-art embedding model designed for precise retrieval across noisy, multilingual, and multimodal data.",match:{equals:"embed-v4.0"},context_window:128e3,prices:{input_mtok:.12}}]},{id:"deepseek",name:"Deepseek",pricing_urls:["https://api-docs.deepseek.com/quick_start/pricing"],api_pattern:"https://api\\.deepseek\\.com",price_comments:'Deepseek off-peak pricing applies "UTC 16:30-00:30" so we switch it around and use the off-peak pricing as the default (first) price then the second price with a constraint is the "standard" pricing that applies "UTC 00:30-16:30".',model_match:{contains:"deepseek"},extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"deepseek-chat",name:"DeepSeek Chat",description:"DeepSeek-V3 is the latest model from the DeepSeek team, building upon the instruction following and coding abilities of the previous versions. Pre-trained on nearly 15 trillion tokens, the reported evaluations reveal that the model outperforms other open-source models and rivals leading closed-source models.",match:{or:[{starts_with:"deepseek-chat"},{equals:"deepseek-chat-v3-0324"}]},context_window:64e3,prices:[{prices:{input_mtok:.135,cache_read_mtok:.035,output_mtok:.55}},{constraint:{start_time:"00:30:00Z",end_time:"16:30:00Z",type:"time_of_date"},prices:{input_mtok:.27,cache_read_mtok:.07,output_mtok:1.1}}]},{id:"deepseek-reasoner",name:"Deepseek R1",description:"DeepSeek R1 is here: Performance on par with OpenAI o1, but open-sourced and with fully open reasoning tokens. It's 671B parameters in size, with 37B active in an inference pass.",match:{or:[{equals:"deepseek-reasoner"},{starts_with:"deepseek-r1"},{equals:"deepseek-r1-0528"}]},context_window:64e3,prices:[{prices:{input_mtok:.135,cache_read_mtok:.035,output_mtok:.55}},{constraint:{start_time:"00:30:00Z",end_time:"16:30:00Z",type:"time_of_date"},prices:{input_mtok:.55,cache_read_mtok:.14,output_mtok:2.19}}]},{id:"deepseek-v4-flash",name:"DeepSeek V4 Flash",description:"DeepSeek-V4-Flash. Supports both non-thinking and thinking (default) modes, JSON output, tool calls, chat prefix completion, and FIM completion (non-thinking only).",match:{or:[{starts_with:"deepseek-v4-flash"}]},context_window:1e6,prices:{input_mtok:.14,cache_read_mtok:.0028,output_mtok:.28}},{id:"deepseek-v4-pro",name:"DeepSeek V4 Pro",description:"DeepSeek-V4-Pro. Supports both non-thinking and thinking (default) modes, JSON output, tool calls, chat prefix completion, and FIM completion (non-thinking only).",match:{or:[{starts_with:"deepseek-v4-pro"}]},context_window:1e6,price_comments:"Standard (non-promotional) pricing. DeepSeek is offering a temporary 75% promotional discount that is not reflected here.",prices:{input_mtok:1.74,cache_read_mtok:.0145,output_mtok:3.48}}]},{id:"fireworks",name:"Fireworks",pricing_urls:["https://fireworks.ai/pricing"],api_pattern:"https://api\\.fireworks\\.ai",model_match:{starts_with:"accounts/fireworks/models/"},extractors:[{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"deepseek-r1-0528",name:"DeepSeek R1 0528",description:"The updated DeepSeek-R1-0528 model delivers major improvements in reasoning, inference, and accuracy through enhanced post-training optimization and greater computational resources. It now performs at a level approaching top-tier models like O3 and Gemini 2.5 Pro, with notable gains in complex tasks such as math and programming.",match:{equals:"accounts/fireworks/models/deepseek-r1-0528"},context_window:16e4,prices:{input_mtok:3,output_mtok:8}},{id:"deepseek-v3-0324",name:"Deepseek V3 03-24",description:"A strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token from Deepseek. Updated checkpoint.",match:{equals:"accounts/fireworks/models/deepseek-v3-0324"},context_window:16e4,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.9,output_mtok:.9}},{id:"deepseek-v3p2",name:"Deepseek V3.2",description:"Model from Deepseek that harmonizes high computational efficiency with superior reasoning and agent performance. 675B parameter MoE model.",match:{equals:"accounts/fireworks/models/deepseek-v3p2"},context_window:163840,prices:{input_mtok:.56,cache_read_mtok:.28,output_mtok:1.68}},{id:"gemma-3-27b-it",name:"Gemma 3 27B Instruct",match:{equals:"accounts/fireworks/models/gemma-3-27b-it"},context_window:131e3,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.1,output_mtok:.1}},{id:"glm-4p7",name:"GLM-4.7",description:"Next-generation general-purpose model from Z.ai optimized for coding, reasoning, and agentic workflows. 352B parameter MoE model with advanced thinking controls.",match:{equals:"accounts/fireworks/models/glm-4p7"},context_window:202752,prices:{input_mtok:.6,output_mtok:2.2}},{id:"gpt-oss-120b",name:"OpenAI gpt-oss-120b",description:"OpenAI's open-weight 117B parameter MoE model designed for production, general purpose, high reasoning use-cases. Features powerful reasoning, agentic tasks, and versatile developer use cases.",match:{equals:"accounts/fireworks/models/gpt-oss-120b"},context_window:131072,prices:{input_mtok:.15,cache_read_mtok:.07,output_mtok:.6}},{id:"gpt-oss-20b",name:"OpenAI gpt-oss-20b",description:"OpenAI's open-weight 21.5B parameter model designed for powerful reasoning, agentic tasks, and versatile developer use cases. Optimized for lower latency and local or specialized tasks.",match:{equals:"accounts/fireworks/models/gpt-oss-20b"},context_window:131072,prices:{input_mtok:.07,cache_read_mtok:.04,output_mtok:.3}},{id:"kimi-k2p5",name:"Kimi K2.5",description:"Moonshot AI's flagship agentic model. Unifies vision and text, thinking and non-thinking modes, and single-agent and multi-agent execution into one model. 1T parameter MoE model.",match:{equals:"accounts/fireworks/models/kimi-k2p5"},context_window:262144,prices:{input_mtok:.6,cache_read_mtok:.1,output_mtok:3}},{id:"llama-v3p1-8b-instruct",name:"Llama 3.1 8B Instruct",description:"The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes. The Llama 3.1 instruction tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.",match:{equals:"accounts/fireworks/models/llama-v3p1-8b-instruct"},context_window:131e3,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.2,output_mtok:.2}},{id:"llama4-maverick-instruct-basic",name:"Llama 4 Maverick Instruct (Basic)",description:"The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes. The Llama 3.1 instruction tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.",match:{equals:"accounts/fireworks/models/llama4-maverick-instruct-basic"},context_window:1e6,prices:{input_mtok:.22,output_mtok:.88}},{id:"minimax-m2p1",name:"MiniMax-M2.1",description:"Built for strong real-world performance across complex, multi-language, and agent-driven workflows. 228B parameter model with robust support for systems, backend, web, mobile, and office-style tasks.",match:{equals:"accounts/fireworks/models/minimax-m2p1"},context_window:204800,prices:{input_mtok:.3,output_mtok:1.2}},{id:"qwen2p5-vl-72b-instruct",name:"Qwen2.5-VL 72B Instruct",description:"Latest Qwen's VLM model",match:{equals:"accounts/fireworks/models/qwen2p5-vl-72b-instruct"},context_window:128e3,price_comments:`docs give just one price - "Pricing Per 1M Tokens", we assume that's input and output`,prices:{input_mtok:.9,output_mtok:.9}},{id:"qwen3-235b-a22b",name:"Qwen3 235B-A22B",description:'Qwen3 is the latest evolution in the Qwen LLM series, featuring both dense and MoE models with major advancements in reasoning, agent capabilities, multilingual support, and instruction following. It uniquely allows seamless switching between "thinking" (for complex logic, math, coding) and "non-thinking" modes (for fast, general dialogue), delivering strong performance across tasks.',match:{equals:"accounts/fireworks/models/qwen3-235b-a22b"},context_window:128e3,prices:{input_mtok:.22,output_mtok:.88}}]},{id:"google",name:"Google",pricing_urls:["https://ai.google.dev/gemini-api/docs/pricing","https://cloud.google.com/vertex-ai/generative-ai/pricing"],api_pattern:"https://(.*\\.)?googleapis\\.com",model_match:{contains:"gemini"},provider_match:{or:[{contains:"google"},{contains:"vertex"},{contains:"gemini"}]},extractors:[{api_flavor:"default",root:"usageMetadata",model_path:"modelVersion",mappings:[{path:"promptTokenCount",dest:"input_tokens",required:!1},{path:"cachedContentTokenCount",dest:"cache_read_tokens",required:!1},{path:["cacheTokensDetails",{type:"array-match",field:"modality",match:{equals:"AUDIO"}},"tokenCount"],dest:"cache_audio_read_tokens",required:!1},{path:["promptTokensDetails",{type:"array-match",field:"modality",match:{equals:"AUDIO"}},"tokenCount"],dest:"input_audio_tokens",required:!1},{path:["candidatesTokensDetails",{type:"array-match",field:"modality",match:{equals:"AUDIO"}},"tokenCount"],dest:"output_audio_tokens",required:!1},{path:"candidatesTokenCount",dest:"output_tokens",required:!1},{path:"thoughtsTokenCount",dest:"output_tokens",required:!1},{path:"toolUsePromptTokenCount",dest:"input_tokens",required:!1}]},{api_flavor:"anthropic",root:"usage",model_path:"model",mappings:[{path:"input_tokens",dest:"input_tokens",required:!0},{path:"cache_creation_input_tokens",dest:"input_tokens",required:!1},{path:"cache_read_input_tokens",dest:"input_tokens",required:!1},{path:"cache_creation_input_tokens",dest:"cache_write_tokens",required:!1},{path:"cache_read_input_tokens",dest:"cache_read_tokens",required:!1},{path:"output_tokens",dest:"output_tokens",required:!0}]},{api_flavor:"chat",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:["prompt_tokens_details","cached_tokens"],dest:"cache_read_tokens",required:!1},{path:["prompt_tokens_details","audio_tokens"],dest:"input_audio_tokens",required:!1},{path:["completion_tokens_details","audio_tokens"],dest:"output_audio_tokens",required:!1},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],fallback_model_providers:["anthropic"],models:[{id:"claude-3-5-haiku",match:{contains:"claude-3-5-haiku"},context_window:2e5,prices:{input_mtok:.8,cache_write_mtok:1,cache_read_mtok:.08,output_mtok:4}},{id:"claude-3-5-sonnet",match:{contains:"claude-3-5-sonnet"},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-7-sonnet",match:{contains:"claude-3-7-sonnet"},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-3-haiku",match:{contains:"claude-3-haiku"},context_window:2e5,prices:{input_mtok:.25,cache_write_mtok:.3,cache_read_mtok:.03,output_mtok:1.25}},{id:"claude-3-opus",match:{contains:"claude-3-opus"},prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-4-opus",match:{or:[{contains:"claude-4-opus"},{contains:"claude-opus-4@"},{contains:"claude-opus-4-0"},{contains:"claude-opus-4-1"},{equals:"claude-opus-4"}]},context_window:2e5,prices:{input_mtok:15,cache_write_mtok:18.75,cache_read_mtok:1.5,output_mtok:75}},{id:"claude-4-sonnet",match:{or:[{contains:"claude-4-sonnet"},{contains:"claude-sonnet-4"}]},context_window:2e5,prices:{input_mtok:3,cache_write_mtok:3.75,cache_read_mtok:.3,output_mtok:15}},{id:"claude-opus-4-6",match:{or:[{contains:"claude-4-6-opus"},{contains:"claude-opus-4-6"},{contains:"claude-4.6-opus"},{contains:"claude-opus-4.6"}]},context_window:2e5,prices:{input_mtok:{base:5,tiers:[{start:2e5,price:10}]},cache_write_mtok:{base:6.25,tiers:[{start:2e5,price:12.5}]},cache_read_mtok:{base:.5,tiers:[{start:2e5,price:1}]},output_mtok:{base:25,tiers:[{start:2e5,price:37.5}]}}},{id:"claude-opus-4-7",match:{or:[{contains:"claude-4-7-opus"},{contains:"claude-opus-4-7"},{contains:"claude-4.7-opus"},{contains:"claude-opus-4.7"}]},context_window:1e6,price_comments:"Flat pricing across full 1M context window. Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models",prices:{input_mtok:5,cache_write_mtok:6.25,cache_read_mtok:.5,output_mtok:25}},{id:"gemini-1.0-pro-vision-001",name:"gemini 1.0 pro vision",description:"Google's first-generation advanced multimodal model that can understand text, code, and images. It provides strong reasoning capabilities and follows instructions effectively.",match:{equals:"gemini-1.0-pro-vision-001"},context_window:32768,price_comments:"I can't find anything about this model or it's pricing, so trusting the original source",prices:{input_mtok:.125,output_mtok:.375}},{id:"gemini-1.5-flash",name:"gemini 1.5 flash",description:"A faster, more cost-effective variant of Gemini 1.5 that maintains strong capabilities while optimizing for performance and cost efficiency. Suitable for production deployments requiring high throughput.",match:{contains:"gemini-1.5-flash"},context_window:1e6,prices:{input_mtok:{base:.075,tiers:[{start:128e3,price:.15}]},cache_read_mtok:{base:.01875,tiers:[{start:128e3,price:.0375}]},output_mtok:{base:.3,tiers:[{start:128e3,price:.6}]}}},{id:"gemini-1.5-pro",name:"gemini 1.5 Pro",description:"Google's most capable multimodal model with an extremely long context window of up to 1 million tokens. It excels at complex reasoning, long-form content processing, and multimodal understanding.",match:{contains:"gemini-1.5-pro"},context_window:1e6,prices:{input_mtok:{base:1.25,tiers:[{start:128e3,price:2.5}]},output_mtok:{base:5,tiers:[{start:128e3,price:10}]}}},{id:"gemini-2.0-flash",name:"gemini 2.0 flash",description:"The newest generation of Google's Gemini models, featuring improved reasoning, instruction following, and factual accuracy, with the Flash variant optimized for cost-efficiency and performance.",match:{or:[{ends_with:"gemini-2.0-flash"},{contains:"gemini-2.0-flash-0"},{contains:"gemini-2.0-flash-exp"},{contains:"gemini-2.0-flash-thinking"},{contains:"gemini-2.0-flash-latest"}]},context_window:1e6,prices:{input_mtok:.1,cache_read_mtok:.025,output_mtok:.4,input_audio_mtok:.7,cache_audio_read_mtok:.175}},{id:"gemini-2.0-flash-lite",name:"gemini 2.0 flash lite",description:"A lighter, more cost-effective version of Gemini 2.0 Flash, designed for applications requiring high efficiency while maintaining good performance. Ideal for high-volume, cost-sensitive deployments.",match:{contains:"gemini-2.0-flash-lite"},context_window:1e6,prices:{input_mtok:.075,output_mtok:.3}},{id:"gemini-2.5-flash",name:"Gemini 2.5 Flash",description:`Gemini 2.5 Flash is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking" capabilities, enabling it to provide responses with greater accuracy and nuanced context handling.`,match:{or:[{equals:"gemini-2.5-flash"},{equals:"gemini-2.5-flash-latest"},{equals:"gemini-2.5-flash-preview-09-2025"}]},prices:{input_mtok:.3,cache_read_mtok:.03,output_mtok:2.5,input_audio_mtok:1,cache_audio_read_mtok:.1}},{id:"gemini-2.5-flash-image",name:"Gemini 2.5 Flash Image",description:"Google's specialized image generation model optimized for fast, high-quality image generation. Outputs images at 1024x1024 resolution, with each image consuming 1290 output tokens.",match:{or:[{equals:"gemini-2.5-flash-image"},{equals:"gemini-2.5-flash-image-preview"}]},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-flash-image. Image output is priced at $30 per 1M tokens, with each 1024x1024 image = 1290 tokens = $0.039/image. Cache pricing is not available for this model.",prices:{input_mtok:.3,output_mtok:30}},{id:"gemini-2.5-flash-lite",name:"Gemini 2.5 Flash Lite",description:'Gemini 2.5 Flash-Lite is a lightweight reasoning model in the Gemini 2.5 family, optimized for ultra-low latency and cost efficiency. It offers improved throughput, faster token generation, and better performance across common benchmarks compared to earlier Flash models. By default, "thinking" (i.e. multi-pass reasoning) is disabled to prioritize speed, but developers can enable it via the Reasoning API parameter to selectively trade off cost for intelligence.',match:{or:[{equals:"gemini-2.5-flash-lite"},{starts_with:"gemini-2.5-flash-lite-preview"}]},context_window:1e6,prices:{input_mtok:.1,cache_read_mtok:.01,output_mtok:.4,input_audio_mtok:.3,cache_audio_read_mtok:.03}},{id:"gemini-2.5-flash-preview",name:"Gemini 2.5 Flash Preview 05-20",description:`Gemini 2.5 Flash May 20th Checkpoint is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking" capabilities, enabling it to provide responses with greater accuracy and nuanced context handling.`,match:{or:[{contains:"gemini-2.5-flash-preview-05-20"},{contains:"gemini-2.5-flash-preview-04-17"},{equals:"gemini-2.5-flash-preview-05-20:thinking"},{equals:"gemini-2.5-flash-preview"},{equals:"gemini-2.5-flash-preview:thinking"}]},price_comments:"from https://cloud.google.com/vertex-ai/generative-ai/pricing should be retired 2025-07-15",prices:{input_mtok:.15,output_mtok:.6},deprecated:!0},{id:"gemini-2.5-pro",name:"Gemini 2.5 Pro",description:`Gemini 2.5 Pro is Google's state-of-the-art AI model designed for advanced reasoning, coding, mathematics, and scientific tasks. It employs "thinking" capabilities, enabling it to reason through responses with enhanced accuracy and nuanced context handling. Gemini 2.5 Pro achieves top-tier performance on multiple benchmarks, including first-place positioning on the LMArena leaderboard, reflecting superior human-preference alignment and complex problem-solving abilities.`,match:{starts_with:"gemini-2.5-pro"},price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro",prices:{input_mtok:{base:1.25,tiers:[{start:2e5,price:2.5}]},cache_read_mtok:{base:.125,tiers:[{start:2e5,price:.25}]},output_mtok:{base:10,tiers:[{start:2e5,price:15}]}}},{id:"gemini-3-flash-preview",name:"Gemini 3 Flash Preview",description:"Google's ultra-fast frontier model optimized for speed and efficiency. Delivers state-of-the-art performance while maintaining low latency and cost, with improved reasoning and coding capabilities.",match:{or:[{equals:"gemini-3-flash-preview"},{starts_with:"gemini-3-flash-preview-"}]},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing. Standard pricing shown; Batch API offers 50% discount on input/output.",prices:{input_mtok:.5,cache_read_mtok:.05,output_mtok:3,input_audio_mtok:1,cache_audio_read_mtok:.1}},{id:"gemini-3-pro-image-preview",name:"Gemini 3 Pro Image Preview",description:"Google's image generation model optimized for high-quality image generation. Supports 1K/2K and 4K resolution outputs with flexible pricing based on image dimensions.",match:{or:[{starts_with:"gemini-3-pro-image-preview"},{equals:"gemini-3-pro-image-preview"}]},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image. Image output is priced at $120 per 1M tokens, with each 1K/2K image = 1120 tokens = $0.134/image and each 4K image = 2000 tokens = $0.24/image.",prices:{input_mtok:2,output_mtok:120}},{id:"gemini-3-pro-preview",name:"Gemini 3 Pro Preview",description:"The best model in the world for multimodal understanding, and our most powerful agentic and vibe-coding model yet.",match:{or:[{starts_with:"gemini-3-pro-preview"},{equals:"gemini-3-pro-text-preview"}]},prices:{input_mtok:{base:2,tiers:[{start:2e5,price:4}]},cache_read_mtok:{base:.2,tiers:[{start:2e5,price:.4}]},output_mtok:{base:12,tiers:[{start:2e5,price:18}]}}},{id:"gemini-3.1-flash-image-preview",name:"Gemini 3.1 Flash Image Preview",description:"Google's latest image generation model (Nano Banana 2) optimized for fast, high-quality image generation. Supports multiple output resolutions from 512px to 4K, with text and thinking output priced separately from image output tokens.",match:{starts_with:"gemini-3.1-flash-image-preview"},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing. Image output is priced at $60 per 1M tokens. Preview model - pricing may change.",prices:{input_mtok:.5,output_mtok:60}},{id:"gemini-3.1-flash-lite",name:"Gemini 3.1 Flash Lite",description:"Google's fastest and most cost-efficient Gemini 3 series model, built for intelligence at scale. Optimized for high-volume, low-latency applications while maintaining strong multimodal capabilities.",match:{starts_with:"gemini-3.1-flash-lite"},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing.",prices:{input_mtok:.25,cache_read_mtok:.025,output_mtok:1.5,input_audio_mtok:.5,cache_audio_read_mtok:.05}},{id:"gemini-3.1-pro-preview",name:"Gemini 3.1 Pro Preview",description:"The latest performance, intelligence, and usability improvements to the best model family in the world for multimodal understanding, agentic capabilities, and vibe-coding.",match:{starts_with:"gemini-3.1-pro-preview"},prices:{input_mtok:{base:2,tiers:[{start:2e5,price:4}]},cache_read_mtok:{base:.2,tiers:[{start:2e5,price:.4}]},output_mtok:{base:12,tiers:[{start:2e5,price:18}]}}},{id:"gemini-3.5-flash",name:"Gemini 3.5 Flash",description:"Google's most intelligent model built for speed, combining frontier intelligence with improved reasoning, coding, and multimodal understanding.",match:{starts_with:"gemini-3.5-flash"},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers offer 50% discount on input/output.",prices:{input_mtok:1.5,cache_read_mtok:.15,output_mtok:9}},{id:"gemini-embedding-001",match:{equals:"gemini-embedding-001"},prices:{input_mtok:.15}},{id:"gemini-flash-1.5",name:"Gemini 1.5 Flash",description:"Gemini 1.5 Flash is a foundation model that performs well at a variety of multimodal tasks such as visual understanding, classification, summarization, and creating content from image, audio and video. It's adept at processing visual and text inputs such as photographs, documents, infographics, and screenshots.",match:{equals:"gemini-flash-1.5"},price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-flash",prices:{input_mtok:{base:.075,tiers:[{start:128e3,price:.15}]},cache_read_mtok:{base:.01875,tiers:[{start:128e3,price:.0375}]},output_mtok:{base:.3,tiers:[{start:128e3,price:.6}]}}},{id:"gemini-flash-1.5-8b",name:"gemini 1.5 flash",description:"A faster, more cost-effective variant of Gemini 1.5 that maintains strong capabilities while optimizing for performance and cost efficiency. Suitable for production deployments requiring high throughput.",match:{equals:"gemini-flash-1.5-8b"},context_window:1e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-flash-8b",prices:{input_mtok:{base:.0375,tiers:[{start:128e3,price:.075}]},cache_read_mtok:{base:.01,tiers:[{start:128e3,price:.02}]},output_mtok:{base:.15,tiers:[{start:128e3,price:.3}]}}},{id:"gemini-live-2.5-flash-preview",match:{or:[{starts_with:"gemini-live-2.5-flash-preview"},{starts_with:"gemini-2.5-flash-native-audio-preview"}]},prices:{input_mtok:.5,output_mtok:2,input_audio_mtok:3,output_audio_mtok:12}},{id:"gemini-pro",name:"gemini 1.0 pro",description:"Google's first-generation advanced multimodal model that can understand text, code, and images. It provides strong reasoning capabilities and follows instructions effectively.",match:{or:[{equals:"gemini-pro"},{equals:"gemini-1.0-pro"}]},context_window:32768,price_comments:"I can't find anything so trusting these prices, not sure the model still exists",prices:{input_mtok:.125,output_mtok:.375}},{id:"gemini-pro-1.5",name:"Gemini 1.5 Pro",description:"Google's latest multimodal model, supports image and video[0] in text or chat prompts.",match:{equals:"gemini-pro-1.5"},context_window:2e6,price_comments:"See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-pro",prices:{input_mtok:{base:1.25,tiers:[{start:128e3,price:2.5}]},cache_read_mtok:{base:.3125,tiers:[{start:128e3,price:.625}]},output_mtok:{base:5,tiers:[{start:128e3,price:10}]}}},{id:"gemma-3",name:"Gemma 3 (free)",description:"Lightweight, state-of the art, open model built from the same technology that powers our Gemini models.",match:{or:[{starts_with:"gemma-3-"},{equals:"gemma-3"}]},prices:{}},{id:"gemma-3n",name:"Gemma 3n (free)",description:"Our open model built for efficient performance on everyday devices like mobile phones, laptops, and tablets.",match:{or:[{starts_with:"gemma-3n"}]},prices:{}}]},{id:"groq",name:"Groq",pricing_urls:["https://groq.com/pricing/"],api_pattern:"https://api\\.groq\\.com",extractors:[{api_flavor:"default",root:"usage",model_path:"model",mappings:[{path:"prompt_tokens",dest:"input_tokens",required:!0},{path:"completion_tokens",dest:"output_tokens",required:!0}]}],models:[{id:"deepseek-r1-distill-llama-70b",name:"DeepSeek R1 Distill Llama 70B",match:{equals:"deepseek-r1-distill-llama-70b"},context_window:131072,prices:{input_mtok:.75,output_mtok:.99}},{id:"gemma-7b-it",match:{equals:"gemma-7b-it"},prices:{input_mtok:.07,output_mtok:.07}},{id:"gemma2-9b-it",name:"Gemma 2 9B 8k",match:{or:[{equals:"gemma2-9b-it"},{equals:"gemma2-9b"}]},prices:{input_mtok:.2,output_mtok:.2}},{id:"llama-3.1-405b-reasoning",match:{equals:"llama-3.1-405b-reasoning"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama-3.1-70b-versatile",match:{equals:"llama-3.1-70b-versatile"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama-3.1-8b-instant",name:"Llama 3.1 8B Instant 128k",match:{equals:"llama-3.1-8b-instant"},prices:{input_mtok:.05,output_mtok:.08}},{id:"llama-3.2-11b-text-preview",match:{equals:"llama-3.2-11b-text-preview"},prices:{input_mtok:.18,output_mtok:.18}},{id:"llama-3.2-11b-vision-preview",match:{equals:"llama-3.2-11b-vision-preview"},prices:{input_mtok:.18,output_mtok:.18}},{id:"llama-3.2-1b-preview",match:{equals:"llama-3.2-1b-preview"},prices:{input_mtok:.04,output_mtok:.04}},{id:"llama-3.2-3b-preview",match:{equals:"llama-3.2-3b-preview"},prices:{input_mtok:.06,output_mtok:.06}},{id:"llama-3.2-90b-text-preview",match:{equals:"llama-3.2-90b-text-preview"},prices:{input_mtok:.9,output_mtok:.9}},{id:"llama-3.2-90b-vision-preview",match:{equals:"llama-3.2-90b-vision-preview"},prices:{input_mtok:.9,output_mtok:.9}},{id:"llama-3.3-70b-specdec",match:{equals:"llama-3.3-70b-specdec"},prices:{input_mtok:.59,output_mtok:.99}},{id:"llama-3.3-70b-versatile",name:"Llama 3.3 70B Versatile 128k",match:{equals:"llama-3.3-70b-versatile"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama-guard-3-8b",match:{equals:"llama-guard-3-8b"},prices:{input_mtok:.2,output_mtok:.2}},{id:"llama2-70b-4096",match:{equals:"llama2-70b-4096"},prices:{input_mtok:.7,output_mtok:.8}},{id:"llama3-70b-8192",match:{equals:"llama3-70b-8192"},prices:{input_mtok:.59,output_mtok:.79}},{id:"llama3-8b-8192",match:{equals:"llama3-8b-8192"},prices:{input_mtok:.05,output_mtok:.08}},{id:"llama3-groq-70b-8192-tool-use-preview",match:{equals:"llama3-groq-70b-8192-tool-use-preview"},prices:{input_mtok:.89,output_mtok:.89}},{id:"llama3-groq-8b-8192-tool-use-preview",match:{equals:"llama3-groq-8b-8192-tool-use-preview"},prices:{input_mtok:.19,output_mtok:.19}},{id:"meta-llama/llama-4-maverick-17b-128e-instruct",name:"Llama 4 Maverick 17B 128E",match:{equals:"meta-llama/llama-4-maverick-17b-128e-instruct"},context_window:131072,prices:{input_mtok:.2,output_mtok:.6}},{id:"meta-llama/llama-4-scout-17b-16e-instruct",name:"Llama 4 Scout (17Bx16E) 128k",match:{equals:"meta-llama/llama-4-scout-17b-16e-instruct"},prices:{input_mtok:.11,output_mtok:.34}},{id:"meta-llama/llama-guard-4-12b",name:"Llama Guard 4 12B",match:{equals:"meta-llama/llama-guard-4-12b"},context_window:131072,prices:{input_mtok:.2,output_mtok:.2}},{id:"mistral-saba-24b",match:{equals:"mistral-saba-24b"},prices:{input_mtok:.79,output_mtok:.79}},{id:"mixtral-8x7b-32768",match:{equals:"mixtral-8x7b-32768"},prices:{input_mtok:.24,output_mtok:.24}},{id:"moonshotai/kimi-k2-instruct",name:"Kimi K2 1T 128k",match:{or:[{equals:"moonshotai/kimi-k2-instruct"},{equals:"moonshotai/kimi-k2-instruct-0905"}]},context_window:131072,prices:{input_mtok:1,cache_read_mtok:.5,output_mtok:3}},{id:"openai/gpt-oss-120b",description:`GPT-OSS 120B is OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with
 120 billion parameters and 128 experts.
 `,match:{or:[{equals:"openai/gpt-oss-120b"},{equals:"openai/gpt-oss-safeguard-20b"}]},context_window:131072,prices:{input_mtok:.15,cache_read_mtok:.075,output_mtok:.6}},{id:"openai/gpt-oss-20b",description:`GPT-OSS 20B is OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with
 20 billion parameters and 32 experts.

package/dist/index.js

@@ -643,7 +643,14 @@ const Q = [
     pricing_urls: ["https://aws.amazon.com/bedrock/pricing/"],
     api_pattern: "https://bedrock-runtime\\.[a-z0-9-]+\\.amazonaws\\.com/",
     provider_match: {
-      contains: "bedrock"
+      or: [
+        {
+          contains: "bedrock"
+        },
+        {
+          contains: "amazon"
+        }
+      ]
     },
     extractors: [
       {
@@ -3567,6 +3574,21 @@ Compared to other leading proprietary and open-weights models Command A delivers
           }
         }
       },
+      {
+        id: "gemini-3.5-flash",
+        name: "Gemini 3.5 Flash",
+        description: "Google's most intelligent model built for speed, combining frontier intelligence with improved reasoning, coding, and multimodal understanding.",
+        match: {
+          starts_with: "gemini-3.5-flash"
+        },
+        context_window: 1e6,
+        price_comments: "See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers offer 50% discount on input/output.",
+        prices: {
+          input_mtok: 1.5,
+          cache_read_mtok: 0.15,
+          output_mtok: 9
+        }
+      },
       {
         id: "gemini-embedding-001",
         match: {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pydantic/genai-prices",
-  "version": "0.0.61",
+  "version": "0.0.62",
   "description": "Calculate prices for calling LLM inference APIs",
   "author": "Pydantic Team",
   "type": "module",