smoltalk 0.4.1 → 0.5.0

package/dist/models.js

@@ -1,4 +1,5 @@
 import { z } from "zod";
+import { mergeModelData, mergeHostedTools, } from "./modelData.js";
 export const providers = [
     "ollama",
     "openai",
@@ -38,6 +39,17 @@ export const textModels = [
         cachedInputTokenCost: 0.075,
         outputTokenCost: 0.6,
         outputTokensPerSecond: 65,
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2023-09",
+        releaseDate: "2024-07-18",
+        lastUpdated: "2024-07-18",
+        family: "gpt-mini",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "openai",
     },
     {
@@ -50,6 +62,17 @@ export const textModels = [
         cachedInputTokenCost: 1.25,
         outputTokenCost: 10,
         outputTokensPerSecond: 143,
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2023-09",
+        releaseDate: "2024-05-13",
+        lastUpdated: "2024-08-06",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "openai",
     },
     {
@@ -69,13 +92,24 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2024-05",
+        releaseDate: "2025-04-16",
+        lastUpdated: "2025-04-16",
+        family: "o",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
         provider: "openai",
     },
     {
         type: "text",
         modelName: "o3-mini",
         description: "o3-mini is our most recent small reasoning model, providing high intelligence at the same cost and latency targets of o1-mini. o3-mini also supports key developer features, like Structured Outputs, function calling, Batch API, and more. Like other models in the o-series, it is designed to excel at science, math, and coding tasks. Knowledge cutoff: June 2024.",
-        maxInputTokens: 500000,
+        maxInputTokens: 200000,
         maxOutputTokens: 100000,
         inputTokenCost: 1.1,
         cachedInputTokenCost: 0.55,
@@ -88,6 +122,17 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text"],
+            output: ["text"],
+        },
+        knowledge: "2024-05",
+        releaseDate: "2024-12-20",
+        lastUpdated: "2025-01-29",
+        family: "o-mini",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
         provider: "openai",
     },
     {
@@ -96,9 +141,9 @@ export const textModels = [
         description: "Latest small o-series model optimized for fast, effective reasoning with exceptional performance in coding and visual tasks. Knowledge cutoff: June 2024.",
         maxInputTokens: 200000,
         maxOutputTokens: 100000,
-        inputTokenCost: 0.6,
-        cachedInputTokenCost: 0.3,
-        outputTokenCost: 2.4,
+        inputTokenCost: 1.1,
+        cachedInputTokenCost: 0.275,
+        outputTokenCost: 4.4,
         outputTokensPerSecond: 135,
         reasoning: {
             levels: ["low", "medium", "high"],
@@ -107,6 +152,17 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-05",
+        releaseDate: "2025-04-16",
+        lastUpdated: "2025-04-16",
+        family: "o-mini",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
         provider: "openai",
     },
     {
@@ -122,7 +178,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-05",
+        releaseDate: "2025-06-10",
+        lastUpdated: "2025-06-10",
+        family: "o-pro",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai-responses",
     },
     {
         type: "text",
@@ -141,6 +208,17 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2023-09",
+        releaseDate: "2024-12-05",
+        lastUpdated: "2024-12-05",
+        family: "o",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
         provider: "openai",
     },
     {
@@ -151,6 +229,17 @@ export const textModels = [
         maxOutputTokens: 4096,
         inputTokenCost: 10,
         outputTokenCost: 30,
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2023-12",
+        releaseDate: "2023-11-06",
+        lastUpdated: "2024-04-09",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: false,
+        temperatureSupported: true,
         disabled: true,
         provider: "openai",
     },
@@ -162,6 +251,17 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 30,
         outputTokenCost: 60,
+        modalities: {
+            input: ["text"],
+            output: ["text"],
+        },
+        knowledge: "2023-11",
+        releaseDate: "2023-11-06",
+        lastUpdated: "2024-04-09",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: false,
+        temperatureSupported: true,
         disabled: true,
         provider: "openai",
     },
@@ -172,7 +272,19 @@ export const textModels = [
         maxInputTokens: 16385,
         maxOutputTokens: 4096,
         inputTokenCost: 0.5,
+        cachedInputTokenCost: 0,
         outputTokenCost: 1.5,
+        modalities: {
+            input: ["text"],
+            output: ["text"],
+        },
+        knowledge: "2021-09-01",
+        releaseDate: "2023-03-01",
+        lastUpdated: "2023-11-06",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: false,
+        temperatureSupported: true,
         disabled: true,
         provider: "openai",
     },
@@ -182,10 +294,21 @@ export const textModels = [
         description: "GPT-4.1 excels at instruction following and tool calling with 1M token context window. Knowledge cutoff: June 2024.",
         maxInputTokens: 1047576,
         maxOutputTokens: 32768,
-        inputTokenCost: 2.0,
+        inputTokenCost: 2,
         cachedInputTokenCost: 0.5,
         outputTokenCost: 8,
         outputTokensPerSecond: 105,
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2024-04",
+        releaseDate: "2025-04-14",
+        lastUpdated: "2025-04-14",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "openai",
     },
     {
@@ -198,6 +321,17 @@ export const textModels = [
         cachedInputTokenCost: 0.1,
         outputTokenCost: 1.6,
         outputTokensPerSecond: 78,
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2024-04",
+        releaseDate: "2025-04-14",
+        lastUpdated: "2025-04-14",
+        family: "gpt-mini",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "openai",
     },
     {
@@ -210,6 +344,17 @@ export const textModels = [
         cachedInputTokenCost: 0.025,
         outputTokenCost: 0.4,
         outputTokensPerSecond: 142,
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-04",
+        releaseDate: "2025-04-14",
+        lastUpdated: "2025-04-14",
+        family: "gpt-nano",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "openai",
     },
     {
@@ -229,7 +374,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-09-30",
+        releaseDate: "2025-08-07",
+        lastUpdated: "2025-08-07",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
@@ -248,7 +404,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-05-30",
+        releaseDate: "2025-08-07",
+        lastUpdated: "2025-08-07",
+        family: "gpt-mini",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
@@ -267,7 +434,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-05-30",
+        releaseDate: "2025-08-07",
+        lastUpdated: "2025-08-07",
+        family: "gpt-nano",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
@@ -285,7 +463,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-09-30",
+        releaseDate: "2025-11-13",
+        lastUpdated: "2025-11-13",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
@@ -304,7 +493,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2025-08-31",
+        releaseDate: "2025-12-11",
+        lastUpdated: "2025-12-11",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
@@ -319,17 +519,34 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2025-08-31",
+        releaseDate: "2025-12-11",
+        lastUpdated: "2025-12-11",
+        family: "gpt-pro",
+        openWeights: false,
+        structuredOutput: false,
+        temperatureSupported: false,
         provider: "openai-responses",
     },
     {
         type: "text",
         modelName: "gpt-5.4",
         description: "GPT-5.4 is the most capable and efficient frontier model for complex professional work. 1M context window, state-of-the-art coding and tool use. Standard pricing for ≤272K tokens, 2x input/1.5x output for >272K. Knowledge cutoff: August 2025.",
-        maxInputTokens: 1_050_000,
+        maxInputTokens: 1050000,
         maxOutputTokens: 128000,
         inputTokenCost: 2.5,
         cachedInputTokenCost: 0.25,
         outputTokenCost: 15,
+        longContext: {
+            inputTokenCost: 5,
+            cachedInputTokenCost: 0.5,
+            outputTokenCost: 22.5,
+            thresholdTokens: 200000,
+        },
         reasoning: {
             levels: ["none", "low", "medium", "high", "xhigh"],
             defaultLevel: "none",
@@ -337,7 +554,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-08-31",
+        releaseDate: "2026-03-05",
+        lastUpdated: "2026-03-05",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
@@ -356,7 +584,18 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2025-08-31",
+        releaseDate: "2026-03-17",
+        lastUpdated: "2026-03-17",
+        family: "gpt-mini",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
@@ -375,16 +614,32 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2025-08-31",
+        releaseDate: "2026-03-17",
+        lastUpdated: "2026-03-17",
+        family: "gpt-nano",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
         modelName: "gpt-5.4-pro",
         description: "GPT-5.4 Pro uses more compute for complex reasoning tasks. 1M context window. Standard pricing for ≤272K tokens. Knowledge cutoff: August 2025.",
-        maxInputTokens: 1_050_000,
+        maxInputTokens: 1050000,
         maxOutputTokens: 128000,
         inputTokenCost: 30,
         outputTokenCost: 180,
+        longContext: {
+            inputTokenCost: 60,
+            outputTokenCost: 270,
+            thresholdTokens: 200000,
+        },
         reasoning: {
             levels: ["medium", "high", "xhigh"],
             defaultLevel: "medium",
@@ -392,18 +647,35 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2025-08-31",
+        releaseDate: "2026-03-05",
+        lastUpdated: "2026-03-05",
+        family: "gpt-pro",
+        openWeights: false,
+        structuredOutput: false,
+        temperatureSupported: false,
         provider: "openai-responses",
     },
     {
         type: "text",
         modelName: "gpt-5.5",
         description: "GPT-5.5 is a new class of intelligence for coding and professional work. 1M context window. Standard pricing for ≤272K tokens, 2x input/1.5x output for >272K. Knowledge cutoff: December 2025.",
-        maxInputTokens: 1_050_000,
+        maxInputTokens: 1050000,
         maxOutputTokens: 128000,
         inputTokenCost: 5,
         cachedInputTokenCost: 0.5,
         outputTokenCost: 30,
         outputTokensPerSecond: 66,
+        longContext: {
+            inputTokenCost: 10,
+            cachedInputTokenCost: 1,
+            outputTokenCost: 45,
+            thresholdTokens: 200000,
+        },
         reasoning: {
             levels: ["none", "low", "medium", "high", "xhigh"],
             defaultLevel: "medium",
@@ -411,16 +683,32 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
-        provider: "openai-responses",
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-12-01",
+        releaseDate: "2026-04-23",
+        lastUpdated: "2026-04-23",
+        family: "gpt",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai",
     },
     {
         type: "text",
         modelName: "gpt-5.5-pro",
         description: "GPT-5.5 Pro uses more compute for complex reasoning tasks. 1M context window. Knowledge cutoff: December 2025.",
-        maxInputTokens: 1_050_000,
+        maxInputTokens: 1050000,
         maxOutputTokens: 128000,
         inputTokenCost: 30,
         outputTokenCost: 180,
+        longContext: {
+            inputTokenCost: 60,
+            outputTokenCost: 270,
+            thresholdTokens: 200000,
+        },
         reasoning: {
             levels: ["none", "low", "medium", "high", "xhigh"],
             defaultLevel: "medium",
@@ -428,17 +716,35 @@ export const textModels = [
             outputsThinking: false,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-12-01",
+        releaseDate: "2026-04-23",
+        lastUpdated: "2026-04-23",
+        family: "gpt-pro",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
         provider: "openai-responses",
     },
     {
         type: "text",
         modelName: "gemini-3.1-pro-preview",
         description: "Latest Gemini 3.1 Pro with 1M context window and 64K output. Standard pricing for ≤200k tokens ($2.00 input/$12.00 output), higher rates for >200k tokens ($4.00 input/$18.00 output). Released Feb 2026.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
-        inputTokenCost: 2.0,
-        outputTokenCost: 12.0,
+        inputTokenCost: 2,
+        cachedInputTokenCost: 0.2,
+        outputTokenCost: 12,
         outputTokensPerSecond: 133,
+        longContext: {
+            inputTokenCost: 4,
+            cachedInputTokenCost: 0.4,
+            outputTokenCost: 18,
+            thresholdTokens: 200000,
+        },
         reasoning: {
             levels: ["low", "medium", "high"],
             defaultLevel: "high",
@@ -446,16 +752,45 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "video", "audio", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2026-02-19",
+        lastUpdated: "2026-02-19",
+        family: "gemini-pro",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-3-pro-preview",
         description: "DEPRECATED: Shut down March 9, 2026. Use gemini-3.1-pro-preview instead.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
-        inputTokenCost: 2.0,
-        outputTokenCost: 12.0,
+        inputTokenCost: 2,
+        cachedInputTokenCost: 0.2,
+        outputTokenCost: 12,
+        longContext: {
+            inputTokenCost: 4,
+            cachedInputTokenCost: 0.4,
+            outputTokenCost: 18,
+            thresholdTokens: 200000,
+        },
+        modalities: {
+            input: ["text", "image", "video", "audio", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2025-11-18",
+        lastUpdated: "2025-11-18",
+        family: "gemini-pro",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         disabled: true,
         provider: "google",
     },
@@ -463,11 +798,12 @@ export const textModels = [
         type: "text",
         modelName: "gemini-3.5-flash",
         description: "Latest Gemini 3.5 Flash model (GA May 2026). Outperforms Gemini 3.1 Pro on coding and agentic suites at 4x the speed. 1M context window, 64K output. Context caching: $0.15/1M read.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
         inputTokenCost: 1.5,
         cachedInputTokenCost: 0.15,
-        outputTokenCost: 9.0,
+        outputTokenCost: 9,
+        inputAudioTokenCost: 1.5,
         reasoning: {
             levels: ["minimal", "low", "medium", "high"],
             defaultLevel: "high",
@@ -475,17 +811,30 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "video", "audio", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2026-05-19",
+        lastUpdated: "2026-05-19",
+        family: "gemini-flash",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-3-flash-preview",
         description: "Gemini 3 Flash preview. Superseded by gemini-3.5-flash. 1M context window and 64K output. Optimized for agentic workflows and coding.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
         inputTokenCost: 0.5,
-        outputTokenCost: 3.0,
+        cachedInputTokenCost: 0.05,
+        outputTokenCost: 3,
         outputTokensPerSecond: 184,
+        inputAudioTokenCost: 1,
         reasoning: {
             levels: ["minimal", "low", "medium", "high"],
             defaultLevel: "high",
@@ -493,17 +842,30 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "video", "audio", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2025-12-17",
+        lastUpdated: "2025-12-17",
+        family: "gemini-flash",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-3.1-flash-lite",
         description: "Most cost-effective Gemini 3.1 model (GA). Thinking support, 1M context window, 64K output. 2.5x faster TTFA and 45% faster output than 2.5 Flash.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
         inputTokenCost: 0.25,
+        cachedInputTokenCost: 0.025,
         outputTokenCost: 1.5,
         outputTokensPerSecond: 379,
+        inputAudioTokenCost: 0.5,
         reasoning: {
             levels: ["minimal", "low", "medium", "high"],
             defaultLevel: "minimal",
@@ -511,79 +873,156 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "video", "audio", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2026-05-07",
+        lastUpdated: "2026-05-07",
+        family: "gemini-flash-lite",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-3.1-flash-lite-preview",
         description: "DEPRECATED: Preview version, discontinued July 9, 2026. Use gemini-3.1-flash-lite instead.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
         inputTokenCost: 0.25,
+        cachedInputTokenCost: 0.025,
         outputTokenCost: 1.5,
+        inputAudioTokenCost: 0.5,
+        modalities: {
+            input: ["text", "image", "video", "audio", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2026-03-03",
+        lastUpdated: "2026-03-03",
+        family: "gemini-flash-lite",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         disabled: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-2.5-pro",
-        description: "High-performance Gemini 2.5 model with 2M context window. Adaptive thinking for complex reasoning and coding. Standard pricing for ≤200k tokens ($1.25 input/$10.00 output), higher rates for >200k tokens ($2.50 input/$15.00 output). Batch API: 50% discount.",
-        maxInputTokens: 2_097_152,
+        description: "High-performance Gemini 2.5 model with 1M context window. Adaptive thinking for complex reasoning and coding. Standard pricing for ≤200k tokens ($1.25 input/$10.00 output), higher rates for >200k tokens ($2.50 input/$15.00 output). Batch API: 50% discount.",
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
         inputTokenCost: 1.25,
-        cachedInputTokenCost: 0.31,
-        outputTokenCost: 10.0,
+        cachedInputTokenCost: 0.125,
+        outputTokenCost: 10,
         outputTokensPerSecond: 145,
+        longContext: {
+            inputTokenCost: 2.5,
+            cachedInputTokenCost: 0.25,
+            outputTokenCost: 15,
+            thresholdTokens: 200000,
+        },
         reasoning: {
             canDisable: false,
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "audio", "video", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2025-06-17",
+        lastUpdated: "2025-06-17",
+        family: "gemini-pro",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-2.5-flash",
         description: "Balanced Gemini 2.5 model with excellent performance-to-cost ratio. Lightning-fast with controllable thinking budgets. 1M context window. Context caching available for up to 75% cost reduction.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
         inputTokenCost: 0.3,
-        cachedInputTokenCost: 0.075,
+        cachedInputTokenCost: 0.03,
         outputTokenCost: 2.5,
         outputTokensPerSecond: 245,
+        inputAudioTokenCost: 1,
         reasoning: {
             canDisable: true,
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "audio", "video", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2025-06-17",
+        lastUpdated: "2025-06-17",
+        family: "gemini-flash",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-2.5-flash-lite",
         description: "Most cost-effective Gemini 2.5 option for high-throughput applications. 1M context window.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 65536,
         inputTokenCost: 0.1,
-        cachedInputTokenCost: 0.025,
+        cachedInputTokenCost: 0.01,
         outputTokenCost: 0.4,
         outputTokensPerSecond: 400,
+        inputAudioTokenCost: 0.3,
         reasoning: {
             canDisable: true,
             outputsThinking: true,
             outputsSignatures: false,
         },
+        modalities: {
+            input: ["text", "image", "audio", "video", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-01",
+        releaseDate: "2025-06-17",
+        lastUpdated: "2025-06-17",
+        family: "gemini-flash-lite",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         provider: "google",
     },
     {
         type: "text",
         modelName: "gemini-2.0-flash",
         description: "Workhorse model for all daily tasks. Strong overall performance and supports real-time streaming Live API. 1M context window. DEPRECATED: Will be shut down on March 31, 2026.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 8192,
         inputTokenCost: 0.1,
+        cachedInputTokenCost: 0.025,
         outputTokenCost: 0.4,
         outputTokensPerSecond: 213,
+        modalities: {
+            input: ["text", "image", "audio", "video", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2024-06",
+        releaseDate: "2024-12-11",
+        lastUpdated: "2024-12-11",
+        family: "gemini-flash",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         disabled: true,
         provider: "google",
     },
@@ -591,7 +1030,7 @@ export const textModels = [
         type: "text",
         modelName: "gemini-2.0-pro-exp-02-05",
         description: "Strongest model quality, especially for code & world knowledge; 2M long context. In private beta.",
-        maxInputTokens: 2_097_152,
+        maxInputTokens: 2097152,
         maxOutputTokens: 8192,
         inputTokenCost: 0.5,
         outputTokenCost: 1.5,
@@ -602,10 +1041,21 @@ export const textModels = [
         type: "text",
         modelName: "gemini-2.0-flash-lite",
         description: "Cost effective offering to support high throughput. DEPRECATED: Will be shut down on March 31, 2026. Use gemini-2.5-flash-lite instead.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 8192,
         inputTokenCost: 0.075,
         outputTokenCost: 0.3,
+        modalities: {
+            input: ["text", "image", "audio", "video", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2024-06",
+        releaseDate: "2024-12-11",
+        lastUpdated: "2024-12-11",
+        family: "gemini-flash-lite",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: true,
         disabled: true,
         provider: "google",
     },
@@ -613,7 +1063,7 @@ export const textModels = [
         type: "text",
         modelName: "gemini-1.5-flash",
         description: "RETIRED: No longer available. Use gemini-2.5-flash instead.",
-        maxInputTokens: 1_048_576,
+        maxInputTokens: 1048576,
         maxOutputTokens: 8192,
         inputTokenCost: 0.01875,
         outputTokenCost: 0.075,
@@ -626,7 +1076,7 @@ export const textModels = [
         type: "text",
         modelName: "gemini-1.5-pro",
         description: "RETIRED: No longer available. Use gemini-2.5-pro instead.",
-        maxInputTokens: 2_097_152,
+        maxInputTokens: 2097152,
         maxOutputTokens: 8192,
         inputTokenCost: 0.3125,
         outputTokenCost: 1.25,
@@ -639,7 +1089,7 @@ export const textModels = [
         type: "text",
         modelName: "gemini-1.0-pro",
         description: "RETIRED: No longer available. Use gemini-2.5-flash instead.",
-        maxInputTokens: 32_760,
+        maxInputTokens: 32760,
         maxOutputTokens: 8192,
         inputTokenCost: 0.125,
         outputTokenCost: 0.375,
@@ -651,10 +1101,11 @@ export const textModels = [
         type: "text",
         modelName: "claude-opus-4-8",
         description: "The most capable Claude model for complex reasoning and agentic coding. Same per-token pricing as Opus 4.7 with improved tool-use efficiency (~290 tokens for tool-use system prompt vs 675 on 4.7). 1M context window, 128K max output.",
-        maxInputTokens: 1_000_000,
-        maxOutputTokens: 128_000,
+        maxInputTokens: 1000000,
+        maxOutputTokens: 128000,
         inputTokenCost: 5,
         cachedInputTokenCost: 0.5,
+        cacheCreationInputTokenCost: 6.25,
         outputTokenCost: 25,
         reasoning: {
             thinkingStyle: "adaptive",
@@ -662,14 +1113,23 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        releaseDate: "2026-05-28",
+        lastUpdated: "2026-05-28",
+        family: "claude-opus",
+        openWeights: false,
+        temperatureSupported: false,
         provider: "anthropic",
     },
     {
         type: "text",
         modelName: "claude-opus-4-7",
         description: "Claude Opus 4.7 for complex reasoning and agentic coding. Features Adaptive Thinking that auto-tunes reasoning compute per request. 1M context window, 128K max output. Knowledge cutoff: January 2026.",
-        maxInputTokens: 1_000_000,
-        maxOutputTokens: 128_000,
+        maxInputTokens: 1000000,
+        maxOutputTokens: 128000,
         inputTokenCost: 5,
         cachedInputTokenCost: 0.5,
         cacheCreationInputTokenCost: 6.25,
@@ -681,14 +1141,24 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2026-01-31",
+        releaseDate: "2026-04-16",
+        lastUpdated: "2026-04-16",
+        family: "claude-opus",
+        openWeights: false,
+        temperatureSupported: false,
         provider: "anthropic",
     },
     {
         type: "text",
         modelName: "claude-opus-4-6",
         description: "Claude Opus 4.6 — legacy model. Use claude-opus-4-7 instead. 1M context window, 128K max output.",
-        maxInputTokens: 1_000_000,
-        maxOutputTokens: 128_000,
+        maxInputTokens: 1000000,
+        maxOutputTokens: 128000,
         inputTokenCost: 5,
         cachedInputTokenCost: 0.5,
         cacheCreationInputTokenCost: 6.25,
@@ -700,14 +1170,24 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-05-31",
+        releaseDate: "2026-02-05",
+        lastUpdated: "2026-03-13",
+        family: "claude-opus",
+        openWeights: false,
+        temperatureSupported: true,
         provider: "anthropic",
     },
     {
         type: "text",
         modelName: "claude-sonnet-4-6",
         description: "The best combination of speed and intelligence. 1M context window, 64K max output.",
-        maxInputTokens: 1_000_000,
-        maxOutputTokens: 64_000,
+        maxInputTokens: 1000000,
+        maxOutputTokens: 64000,
         inputTokenCost: 3,
         cachedInputTokenCost: 0.3,
         cacheCreationInputTokenCost: 3.75,
@@ -719,14 +1199,24 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-08-31",
+        releaseDate: "2026-02-17",
+        lastUpdated: "2026-03-13",
+        family: "claude-sonnet",
+        openWeights: false,
+        temperatureSupported: true,
         provider: "anthropic",
     },
     {
         type: "text",
         modelName: "claude-haiku-4-5-20251001",
         description: "The fastest Claude model with near-frontier intelligence. 200K context window, 64K max output.",
-        maxInputTokens: 200_000,
-        maxOutputTokens: 64_000,
+        maxInputTokens: 200000,
+        maxOutputTokens: 64000,
         inputTokenCost: 1,
         cachedInputTokenCost: 0.1,
         cacheCreationInputTokenCost: 1.25,
@@ -738,13 +1228,23 @@ export const textModels = [
             outputsThinking: true,
             outputsSignatures: true,
         },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-02-28",
+        releaseDate: "2025-10-15",
+        lastUpdated: "2025-10-15",
+        family: "claude-haiku",
+        openWeights: false,
+        temperatureSupported: true,
         provider: "anthropic",
     },
     {
         type: "text",
         modelName: "claude-3-7-sonnet-latest",
         description: "Claude 3.7 Sonnet — legacy model. Use claude-sonnet-4-6 instead.",
-        maxInputTokens: 200_000,
+        maxInputTokens: 200000,
         maxOutputTokens: 8192,
         inputTokenCost: 3,
         outputTokenCost: 15,
@@ -762,7 +1262,7 @@ export const textModels = [
         type: "text",
         modelName: "claude-3-5-haiku-latest",
         description: "Claude 3.5 Haiku — legacy model. Use claude-haiku-4-5-20251001 instead.",
-        maxInputTokens: 200_000,
+        maxInputTokens: 200000,
         maxOutputTokens: 8192,
         inputTokenCost: 0.8,
         outputTokenCost: 4,
@@ -770,6 +1270,116 @@ export const textModels = [
         disabled: true,
         provider: "anthropic",
     },
+    {
+        type: "text",
+        modelName: "claude-haiku-4-5",
+        description: "Undated alias for claude-haiku-4-5-20251001. The fastest Claude model with near-frontier intelligence.",
+        maxInputTokens: 200000,
+        maxOutputTokens: 64000,
+        inputTokenCost: 1,
+        cachedInputTokenCost: 0.1,
+        cacheCreationInputTokenCost: 1.25,
+        outputTokenCost: 5,
+        reasoning: {
+            thinkingStyle: "budget",
+            canDisable: true,
+            outputsThinking: true,
+            outputsSignatures: true,
+        },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-02-28",
+        releaseDate: "2025-10-15",
+        lastUpdated: "2025-10-15",
+        family: "claude-haiku",
+        openWeights: false,
+        temperatureSupported: true,
+        provider: "anthropic",
+    },
+    {
+        type: "text",
+        modelName: "claude-opus-4-5",
+        description: "Claude Opus 4.5 — earlier Opus generation. Prefer claude-opus-4-8.",
+        maxInputTokens: 200000,
+        maxOutputTokens: 64000,
+        inputTokenCost: 5,
+        cachedInputTokenCost: 0.5,
+        cacheCreationInputTokenCost: 6.25,
+        outputTokenCost: 25,
+        reasoning: {
+            thinkingStyle: "budget",
+            canDisable: true,
+            outputsThinking: true,
+            outputsSignatures: true,
+        },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-03-31",
+        releaseDate: "2025-11-24",
+        lastUpdated: "2025-11-24",
+        family: "claude-opus",
+        openWeights: false,
+        temperatureSupported: true,
+        provider: "anthropic",
+    },
+    {
+        type: "text",
+        modelName: "claude-sonnet-4-5",
+        description: "Claude Sonnet 4.5 — earlier Sonnet generation. Prefer claude-sonnet-4-6.",
+        maxInputTokens: 200000,
+        maxOutputTokens: 64000,
+        inputTokenCost: 3,
+        cachedInputTokenCost: 0.3,
+        cacheCreationInputTokenCost: 3.75,
+        outputTokenCost: 15,
+        reasoning: {
+            thinkingStyle: "budget",
+            canDisable: true,
+            outputsThinking: true,
+            outputsSignatures: true,
+        },
+        modalities: {
+            input: ["text", "image", "pdf"],
+            output: ["text"],
+        },
+        knowledge: "2025-07-31",
+        releaseDate: "2025-09-29",
+        lastUpdated: "2025-09-29",
+        family: "claude-sonnet",
+        openWeights: false,
+        temperatureSupported: true,
+        provider: "anthropic",
+    },
+    {
+        type: "text",
+        modelName: "gpt-5-pro",
+        description: "GPT-5 Pro uses more compute for complex reasoning tasks. Available via the Responses API.",
+        maxInputTokens: 400000,
+        maxOutputTokens: 272000,
+        inputTokenCost: 15,
+        outputTokenCost: 120,
+        reasoning: {
+            canDisable: false,
+            outputsThinking: false,
+            outputsSignatures: false,
+        },
+        modalities: {
+            input: ["text", "image"],
+            output: ["text"],
+        },
+        knowledge: "2024-09-30",
+        releaseDate: "2025-10-06",
+        lastUpdated: "2025-10-06",
+        family: "gpt-pro",
+        openWeights: false,
+        structuredOutput: true,
+        temperatureSupported: false,
+        provider: "openai-responses",
+    },
 ];
 export const imageModels = [
     {
@@ -847,19 +1457,185 @@ export const embeddingsModels = [
         tokenCost: 0.2,
     },
 ];
+export const hostedTools = [
+    {
+        name: "web_search",
+        provider: "anthropic",
+        category: "web_search",
+        description: "Server-side web search with citations.",
+        providerToolId: "web_search",
+        pricing: { unit: "per_call", amount: 0.01, note: "$10 per 1,000 searches, plus content tokens." },
+    },
+    {
+        name: "web_fetch",
+        provider: "anthropic",
+        category: "url_context",
+        description: "Fetch and read a specific URL.",
+        providerToolId: "web_fetch",
+        pricing: { unit: "free", note: "No extra charge beyond tokens." },
+    },
+    {
+        name: "code_execution",
+        provider: "anthropic",
+        category: "code_execution",
+        description: "Run code in a sandboxed container.",
+        providerToolId: "code_execution",
+        pricing: { unit: "per_hour", amount: 0.05, freeAllowance: "50 container-hours/day", note: "Free when used with web_search or web_fetch." },
+    },
+    {
+        name: "web_search",
+        provider: "openai-responses",
+        category: "web_search",
+        description: "Server-side web search (Responses API).",
+        providerToolId: "web_search",
+        pricing: { unit: "per_call", amount: 0.01, note: "$10/1k standard; $25/1k preview non-reasoning; plus ~8k input tokens per call. Varies by search_context_size." },
+    },
+    {
+        name: "file_search",
+        provider: "openai-responses",
+        category: "file_search",
+        description: "Semantic + keyword search over uploaded files.",
+        providerToolId: "file_search",
+        pricing: { unit: "per_call", amount: 0.0025, note: "$2.50 per 1,000 calls, plus $0.10/GB-day vector storage." },
+    },
+    {
+        name: "code_interpreter",
+        provider: "openai-responses",
+        category: "code_execution",
+        description: "Run Python in a sandboxed container.",
+        providerToolId: "code_interpreter",
+        pricing: { unit: "per_session", amount: 0.03, note: "Per container session." },
+    },
+    {
+        name: "image_generation",
+        provider: "openai-responses",
+        category: "image_generation",
+        description: "Generate images with gpt-image-1 as a tool.",
+        providerToolId: "image_generation",
+        pricing: { unit: "tokens", note: "gpt-image-1 token pricing: $5/$10/$40 per 1M text-in/image-in/image-out." },
+    },
+    {
+        name: "google_search",
+        provider: "google",
+        category: "web_search",
+        description: "Grounding with Google Search.",
+        providerToolId: "google_search",
+        pricing: {
+            unit: "per_call",
+            amount: 0.014,
+            freeAllowance: "5,000 grounded prompts/month (Gemini 3)",
+            note: "Billed per query; a request may issue multiple queries.",
+            perModel: {
+                "gemini-2.5-pro": { amount: 0.035, freeAllowance: "1,500/day shared", note: "Billed per prompt (Gemini 2.5)." },
+                "gemini-2.5-flash": { amount: 0.035, freeAllowance: "1,500/day shared", note: "Billed per prompt (Gemini 2.5)." },
+                "gemini-2.5-flash-lite": { amount: 0.035, freeAllowance: "1,500/day shared", note: "Billed per prompt (Gemini 2.5)." },
+            },
+        },
+    },
+    {
+        name: "code_execution",
+        provider: "google",
+        category: "code_execution",
+        description: "Run Python generated by the model.",
+        providerToolId: "code_execution",
+        pricing: { unit: "tokens", note: "Billed as tokens; no separate tool fee." },
+    },
+    {
+        name: "url_context",
+        provider: "google",
+        category: "url_context",
+        description: "Ground responses on specific URLs you provide.",
+        providerToolId: "url_context",
+        pricing: { unit: "tokens", note: "Tokens only." },
+    },
+    {
+        name: "maps_grounding",
+        provider: "google",
+        category: "maps_grounding",
+        description: "Grounding with Google Maps (Gemini 3 only).",
+        providerToolId: "google_maps",
+        models: ["gemini-3-pro-preview", "gemini-3.1-pro-preview", "gemini-3-flash-preview", "gemini-3.5-flash", "gemini-3.1-flash-lite"],
+        pricing: { unit: "per_call", note: "Gemini 3 family only; see Google pricing." },
+    },
+];
 export const registeredTextModels = [];
 export function registerTextModel(model) {
     registeredTextModels.push({ ...model, type: "text" });
 }
-export function getModel(modelName) {
-    const allModels = [
+let registeredModelData = null;
+export function registerModelData(blob) {
+    registeredModelData = blob;
+}
+export function clearModelData() {
+    registeredModelData = null;
+}
+export function getRegisteredModelData() {
+    return registeredModelData;
+}
+function baselineModels() {
+    return [
         ...textModels,
         ...imageModels,
         ...speechToTextModels,
         ...registeredTextModels,
         ...embeddingsModels,
     ];
-    return allModels.find((model) => model.modelName === modelName);
+}
+export function getModel(modelName, requestData) {
+    let models = baselineModels();
+    if (registeredModelData) {
+        models = mergeModelData(models, registeredModelData.models);
+    }
+    if (requestData) {
+        models = mergeModelData(models, requestData.models);
+    }
+    return models.find((model) => model.modelName === modelName);
+}
+export function getHostedTools(opts = {}) {
+    // Start from a copy so callers can never mutate the baseline registry.
+    let tools = [...hostedTools];
+    if (registeredModelData) {
+        tools = mergeHostedTools(tools, registeredModelData.hostedTools);
+    }
+    if (opts.modelData) {
+        tools = mergeHostedTools(tools, opts.modelData.hostedTools);
+    }
+    let modelProvider;
+    if (opts.model) {
+        modelProvider = getModel(opts.model, opts.modelData)?.provider;
+    }
+    return tools.filter((tool) => {
+        if (tool.disabled && !opts.includeDisabled) {
+            return false;
+        }
+        if (opts.provider && tool.provider !== opts.provider) {
+            return false;
+        }
+        if (opts.category && tool.category !== opts.category) {
+            return false;
+        }
+        if (opts.model) {
+            if (tool.provider !== modelProvider) {
+                return false;
+            }
+            if (tool.models && !tool.models.includes(opts.model)) {
+                return false;
+            }
+        }
+        return true;
+    });
+}
+export function hostedToolPricingFor(tool, model) {
+    if (!tool.pricing) {
+        return undefined;
+    }
+    // Drop perModel from the base; merge the override for `model` when present.
+    // The return type is HostedToolPrice, so the result can never carry perModel.
+    const { perModel, ...base } = tool.pricing;
+    if (model && perModel && perModel[model]) {
+        return { ...base, ...perModel[model] };
+    }
+    return base;
 }
 export function isImageModel(model) {
     return model.type === "image";