@blockrun/clawrouter 0.12.78 → 0.12.80

package/README.md

@@ -96,7 +96,6 @@ Choose your routing strategy with `/model <profile>`:
 | `/model auto`    | Balanced (default) | 74-100% | General use      |
 | `/model eco`     | Cheapest possible  | 95-100% | Maximum savings  |
 | `/model premium` | Best quality       | 0%      | Mission-critical |
-| `/model free`    | 11 free NVIDIA models | 100% | Zero cost        |
 
 **Shortcuts:** `/model grok`, `/model br-sonnet`, `/model gpt5`, `/model o3`
 
@@ -110,12 +109,12 @@ Choose your routing strategy with `/model <profile>`:
 Request → Weighted Scorer (15 dimensions) → Tier → Best Model → Response
 ```
 
-| Tier      | ECO Model                              | AUTO Model                            | PREMIUM Model                |
-| --------- | -------------------------------------- | ------------------------------------- | ---------------------------- |
-| SIMPLE    | nvidia/gpt-oss-120b (**FREE**)         | gemini-2.5-flash ($0.30/$2.50)        | kimi-k2.5                    |
-| MEDIUM    | nvidia/deepseek-v3.2 (**FREE**)        | kimi-k2.5 ($0.60/$3.00)              | gpt-5.3-codex ($1.75/$14.00) |
-| COMPLEX   | nvidia/nemotron-ultra-253b (**FREE**)  | gemini-3.1-pro ($2/$12)               | claude-opus-4.6 ($5/$25)     |
-| REASONING | grok-4-1-fast ($0.20/$0.50)            | grok-4-1-fast-reasoning ($0.20/$0.50) | claude-sonnet-4.6 ($3/$15)   |
+| Tier      | ECO Model                           | AUTO Model                            | PREMIUM Model                |
+| --------- | ----------------------------------- | ------------------------------------- | ---------------------------- |
+| SIMPLE    | nvidia/gpt-oss-120b (**FREE**)      | gemini-2.5-flash ($0.30/$2.50)        | kimi-k2.5                    |
+| MEDIUM    | gemini-3.1-flash-lite ($0.25/$1.50) | kimi-k2.5 ($0.60/$3.00)               | gpt-5.3-codex ($1.75/$14.00) |
+| COMPLEX   | gemini-3.1-flash-lite ($0.25/$1.50) | gemini-3.1-pro ($2/$12)               | claude-opus-4.6 ($5/$25)     |
+| REASONING | grok-4-1-fast ($0.20/$0.50)         | grok-4-1-fast-reasoning ($0.20/$0.50) | claude-sonnet-4.6 ($3/$15)   |
 
 **Blended average: $2.05/M** vs $25/M for Claude Opus = **92% savings**
 
@@ -234,7 +233,7 @@ Edit existing images with `/img2img`:
 | openai/gpt-5.2-pro          |    $21.00 |    $168.00 |    $0.0945 | 400K    | reasoning, tools                  |
 | openai/gpt-5.4-pro          |    $30.00 |    $180.00 |    $0.1050 | 400K    | reasoning, tools                  |
 
-> **Free tier:** 11 NVIDIA models cost nothing — use `/model free` for smart routing across all of them, or pick one directly (e.g., `/model nemotron`). Auto-fallback when wallet is empty also uses free models matched to task complexity.
+> **Free tier:** 11 NVIDIA models cost nothing — `/model free` points to nemotron-ultra-253b, or pick any free model directly (e.g., `/model nemotron`, `/model deepseek-free`, `/model devstral`).
 > **Best value:** `gpt-5-nano` and `gemini-2.5-flash-lite` deliver strong results at ~$0.0003/request.
 
 ---
@@ -465,7 +464,7 @@ ClawRouter integrates with OpenClaw (Claude Code), ElizaOS, and any agent that m
 
 ### Is ClawRouter free?
 
-ClawRouter itself is free and MIT licensed. You pay only for the LLM API calls routed through it — and 11 NVIDIA models (DeepSeek V3.2, Nemotron Ultra 253B, Mistral Large 675B, Llama 4 Maverick, and more) are completely free. Use `/model free` for zero-cost smart routing.
+ClawRouter itself is free and MIT licensed. You pay only for the LLM API calls routed through it — and 11 NVIDIA models (DeepSeek V3.2, Nemotron Ultra 253B, Mistral Large 675B, Llama 4 Maverick, and more) are completely free. Use `/model free` for Nemotron Ultra 253B, or pick any free model by name.
 
 ---
 

package/dist/cli.js

@@ -39086,11 +39086,7 @@ var RulesStrategy = class {
     let tierConfigs;
     let profileSuffix;
     let profile;
-    if (routingProfile === "free" && config.freeTiers) {
-      tierConfigs = config.freeTiers;
-      profileSuffix = " | free";
-      profile = "free";
-    } else if (routingProfile === "eco" && config.ecoTiers) {
+    if (routingProfile === "eco" && config.ecoTiers) {
       tierConfigs = config.ecoTiers;
       profileSuffix = " | eco";
       profile = "eco";
@@ -40218,7 +40214,7 @@ var DEFAULT_ROUTING_CONFIG = {
         // $0.20/$1.25, 1M context
         "xai/grok-4-fast-non-reasoning",
         // 1,143ms, $0.20/$0.50 — fast fallback
-        "nvidia/gpt-oss-120b"
+        "free/gpt-oss-120b"
         // 1,252ms, FREE fallback
       ]
     },
@@ -40279,13 +40275,13 @@ var DEFAULT_ROUTING_CONFIG = {
       ]
     }
   },
-  // Eco tier configs - absolute cheapest, free-first (blockrun/eco)
+  // Eco tier configs - absolute cheapest (blockrun/eco)
   ecoTiers: {
     SIMPLE: {
-      primary: "nvidia/gpt-oss-120b",
+      primary: "free/gpt-oss-120b",
       // FREE! $0.00/$0.00
       fallback: [
-        "nvidia/gpt-oss-20b",
+        "free/gpt-oss-20b",
         // FREE — smaller, faster
         "google/gemini-3.1-flash-lite",
         // $0.25/$1.50 — newest flash-lite
@@ -40298,13 +40294,9 @@ var DEFAULT_ROUTING_CONFIG = {
       ]
     },
     MEDIUM: {
-      primary: "nvidia/deepseek-v3.2",
-      // FREE — DeepSeek V3.2 quality at zero cost
+      primary: "google/gemini-3.1-flash-lite",
+      // $0.25/$1.50 — newest flash-lite
       fallback: [
-        "nvidia/gpt-oss-120b",
-        // FREE fallback
-        "google/gemini-3.1-flash-lite",
-        // $0.25/$1.50
         "openai/gpt-5.4-nano",
         // $0.20/$1.25
         "google/gemini-2.5-flash-lite",
@@ -40314,15 +40306,9 @@ var DEFAULT_ROUTING_CONFIG = {
       ]
     },
     COMPLEX: {
-      primary: "nvidia/nemotron-ultra-253b",
-      // FREE — 253B reasoning model
+      primary: "google/gemini-3.1-flash-lite",
+      // $0.25/$1.50
       fallback: [
-        "nvidia/mistral-large-3-675b",
-        // FREE — 675B brute-force
-        "nvidia/deepseek-v3.2",
-        // FREE
-        "google/gemini-3.1-flash-lite",
-        // $0.25/$1.50
         "google/gemini-2.5-flash-lite",
         "xai/grok-4-0709",
         "google/gemini-2.5-flash",
@@ -40332,12 +40318,7 @@ var DEFAULT_ROUTING_CONFIG = {
     REASONING: {
       primary: "xai/grok-4-1-fast-reasoning",
       // $0.20/$0.50
-      fallback: [
-        "xai/grok-4-fast-reasoning",
-        "nvidia/nemotron-ultra-253b",
-        // FREE reasoning fallback
-        "deepseek/deepseek-reasoner"
-      ]
+      fallback: ["xai/grok-4-fast-reasoning", "deepseek/deepseek-reasoner"]
     }
   },
   // Premium tier configs - best quality (blockrun/premium)
@@ -40451,73 +40432,6 @@ var DEFAULT_ROUTING_CONFIG = {
       ]
     }
   },
-  // Free tier configs - NVIDIA free models, smart-routed by task type (blockrun/free)
-  freeTiers: {
-    SIMPLE: {
-      primary: "nvidia/gpt-oss-20b",
-      // Fastest: small 20B for simple tasks
-      fallback: [
-        "nvidia/gpt-oss-120b",
-        // Solid general-purpose
-        "nvidia/nemotron-super-49b",
-        // Thinking mode
-        "nvidia/llama-4-maverick",
-        // MoE broad coverage
-        "nvidia/glm-4.7"
-        // Thinking mode
-      ]
-    },
-    MEDIUM: {
-      primary: "nvidia/deepseek-v3.2",
-      // DeepSeek V3.2 quality, zero cost
-      fallback: [
-        "nvidia/gpt-oss-120b",
-        // Strong 120B general-purpose
-        "nvidia/nemotron-super-49b",
-        // Thinking mode
-        "nvidia/mistral-large-3-675b",
-        // Largest Mistral
-        "nvidia/llama-4-maverick",
-        // MoE breadth
-        "nvidia/glm-4.7"
-        // Thinking mode
-      ]
-    },
-    COMPLEX: {
-      primary: "nvidia/nemotron-ultra-253b",
-      // Strongest free: 253B reasoning
-      fallback: [
-        "nvidia/mistral-large-3-675b",
-        // 675B massive params
-        "nvidia/deepseek-v3.2",
-        // V3.2 quality
-        "nvidia/nemotron-3-super-120b",
-        // Thinking mode MoE
-        "nvidia/qwen3-coder-480b",
-        // 480B MoE for code-heavy tasks
-        "nvidia/devstral-2-123b",
-        // Coding-focused
-        "nvidia/gpt-oss-120b"
-        // Last resort
-      ]
-    },
-    REASONING: {
-      primary: "nvidia/nemotron-ultra-253b",
-      // Best free reasoning: 253B
-      fallback: [
-        "nvidia/nemotron-3-super-120b",
-        // Thinking mode MoE
-        "nvidia/nemotron-super-49b",
-        // Thinking mode
-        "nvidia/deepseek-v3.2",
-        // DeepSeek reasoning
-        "nvidia/mistral-large-3-675b",
-        // Brute-force params
-        "nvidia/glm-4.7"
-        // GLM thinking mode
-      ]
-    }
-  },
   overrides: {
     maxTokensForceComplex: 1e5,
     structuredOutputMinTier: "MEDIUM",
@@ -40597,27 +40511,38 @@ var MODEL_ALIASES = {
   // delisted 2026-03-12
   "xai/grok-3-fast": "xai/grok-4-fast-reasoning",
   // delisted (too expensive)
-  // NVIDIA — existing alias kept for backward compat
-  nvidia: "nvidia/gpt-oss-120b",
-  "gpt-120b": "nvidia/gpt-oss-120b",
-  "gpt-20b": "nvidia/gpt-oss-20b",
-  // Free model aliases — "-free" suffix for models with paid twins
-  "deepseek-free": "nvidia/deepseek-v3.2",
-  "mistral-free": "nvidia/mistral-large-3-675b",
-  "glm-free": "nvidia/glm-4.7",
-  "llama-free": "nvidia/llama-4-maverick",
-  // Bare-name aliases for unique free models
-  nemotron: "nvidia/nemotron-ultra-253b",
-  "nemotron-ultra": "nvidia/nemotron-ultra-253b",
-  "nemotron-253b": "nvidia/nemotron-ultra-253b",
-  "nemotron-super": "nvidia/nemotron-super-49b",
-  "nemotron-49b": "nvidia/nemotron-super-49b",
-  "nemotron-120b": "nvidia/nemotron-3-super-120b",
-  devstral: "nvidia/devstral-2-123b",
-  "devstral-2": "nvidia/devstral-2-123b",
-  "qwen-coder": "nvidia/qwen3-coder-480b",
-  "qwen-coder-free": "nvidia/qwen3-coder-480b",
-  maverick: "nvidia/llama-4-maverick",
+  // NVIDIA — backward compat aliases (nvidia/xxx → free/xxx)
+  nvidia: "free/gpt-oss-120b",
+  "gpt-120b": "free/gpt-oss-120b",
+  "gpt-20b": "free/gpt-oss-20b",
+  "nvidia/gpt-oss-120b": "free/gpt-oss-120b",
+  "nvidia/gpt-oss-20b": "free/gpt-oss-20b",
+  "nvidia/nemotron-ultra-253b": "free/nemotron-ultra-253b",
+  "nvidia/nemotron-3-super-120b": "free/nemotron-3-super-120b",
+  "nvidia/nemotron-super-49b": "free/nemotron-super-49b",
+  "nvidia/deepseek-v3.2": "free/deepseek-v3.2",
+  "nvidia/mistral-large-3-675b": "free/mistral-large-3-675b",
+  "nvidia/qwen3-coder-480b": "free/qwen3-coder-480b",
+  "nvidia/devstral-2-123b": "free/devstral-2-123b",
+  "nvidia/glm-4.7": "free/glm-4.7",
+  "nvidia/llama-4-maverick": "free/llama-4-maverick",
+  // Free model shorthand aliases
+  "deepseek-free": "free/deepseek-v3.2",
+  "mistral-free": "free/mistral-large-3-675b",
+  "glm-free": "free/glm-4.7",
+  "llama-free": "free/llama-4-maverick",
+  nemotron: "free/nemotron-ultra-253b",
+  "nemotron-ultra": "free/nemotron-ultra-253b",
+  "nemotron-253b": "free/nemotron-ultra-253b",
+  "nemotron-super": "free/nemotron-super-49b",
+  "nemotron-49b": "free/nemotron-super-49b",
+  "nemotron-120b": "free/nemotron-3-super-120b",
+  devstral: "free/devstral-2-123b",
+  "devstral-2": "free/devstral-2-123b",
+  "qwen-coder": "free/qwen3-coder-480b",
+  "qwen-coder-free": "free/qwen3-coder-480b",
+  maverick: "free/llama-4-maverick",
+  free: "free/nemotron-ultra-253b",
   // MiniMax
   minimax: "minimax/minimax-m2.7",
   "minimax-m2.7": "minimax/minimax-m2.7",
@@ -40629,7 +40554,7 @@ var MODEL_ALIASES = {
   // Routing profile aliases (common variations)
   "auto-router": "auto",
   router: "auto"
-  // Note: auto, free, eco, premium are virtual routing profiles registered in BLOCKRUN_MODELS
+  // Note: auto, eco, premium are virtual routing profiles registered in BLOCKRUN_MODELS
   // They don't need aliases since they're already top-level model IDs
 };
 function resolveModelAlias(model) {
@@ -40662,14 +40587,6 @@ var BLOCKRUN_MODELS = [
     contextWindow: 105e4,
     maxOutput: 128e3
   },
-  {
-    id: "free",
-    name: "Free (Smart Router - 11 NVIDIA Models)",
-    inputPrice: 0,
-    outputPrice: 0,
-    contextWindow: 131072,
-    maxOutput: 16384
-  },
   {
     id: "eco",
     name: "Eco (Smart Router - Cost Optimized)",
@@ -41173,11 +41090,12 @@ var BLOCKRUN_MODELS = [
     agentic: true,
     toolCalling: true
   },
-  // NVIDIA - Free models (hosted by NVIDIA, billingMode: "free" on server)
-  // toolCalling intentionally omitted on all free models: structured function
-  // calling support unverified. Excluded from tool-heavy routing paths.
+  // Free models (hosted by NVIDIA, billingMode: "free" on server)
+  // IDs use "free/" prefix so users see them as free in the /model picker.
+  // ClawRouter maps free/xxx → nvidia/xxx before sending to BlockRun upstream.
+  // toolCalling intentionally omitted: structured function calling unverified.
   {
-    id: "nvidia/gpt-oss-120b",
+    id: "free/gpt-oss-120b",
     name: "[Free] GPT-OSS 120B",
     version: "120b",
     inputPrice: 0,
@@ -41186,7 +41104,7 @@ var BLOCKRUN_MODELS = [
     maxOutput: 16384
   },
   {
-    id: "nvidia/gpt-oss-20b",
+    id: "free/gpt-oss-20b",
     name: "[Free] GPT-OSS 20B",
     version: "20b",
     inputPrice: 0,
@@ -41195,7 +41113,7 @@ var BLOCKRUN_MODELS = [
     maxOutput: 16384
   },
   {
-    id: "nvidia/nemotron-ultra-253b",
+    id: "free/nemotron-ultra-253b",
     name: "[Free] Nemotron Ultra 253B",
     version: "253b",
     inputPrice: 0,
@@ -41205,7 +41123,7 @@ var BLOCKRUN_MODELS = [
     reasoning: true
   },
   {
-    id: "nvidia/nemotron-3-super-120b",
+    id: "free/nemotron-3-super-120b",
     name: "[Free] Nemotron 3 Super 120B",
     version: "3-super-120b",
     inputPrice: 0,
@@ -41215,7 +41133,7 @@ var BLOCKRUN_MODELS = [
     reasoning: true
   },
   {
-    id: "nvidia/nemotron-super-49b",
+    id: "free/nemotron-super-49b",
     name: "[Free] Nemotron Super 49B",
     version: "super-49b",
     inputPrice: 0,
@@ -41225,7 +41143,7 @@ var BLOCKRUN_MODELS = [
     reasoning: true
   },
   {
-    id: "nvidia/deepseek-v3.2",
+    id: "free/deepseek-v3.2",
     name: "[Free] DeepSeek V3.2",
     version: "v3.2",
     inputPrice: 0,
@@ -41235,7 +41153,7 @@ var BLOCKRUN_MODELS = [
     reasoning: true
   },
   {
-    id: "nvidia/mistral-large-3-675b",
+    id: "free/mistral-large-3-675b",
     name: "[Free] Mistral Large 675B",
     version: "3-675b",
     inputPrice: 0,
@@ -41245,7 +41163,7 @@ var BLOCKRUN_MODELS = [
     reasoning: true
   },
   {
-    id: "nvidia/qwen3-coder-480b",
+    id: "free/qwen3-coder-480b",
     name: "[Free] Qwen3 Coder 480B",
     version: "480b",
     inputPrice: 0,
@@ -41254,7 +41172,7 @@ var BLOCKRUN_MODELS = [
     maxOutput: 16384
   },
   {
-    id: "nvidia/devstral-2-123b",
+    id: "free/devstral-2-123b",
     name: "[Free] Devstral 2 123B",
     version: "2-123b",
     inputPrice: 0,
@@ -41263,7 +41181,7 @@ var BLOCKRUN_MODELS = [
     maxOutput: 16384
   },
   {
-    id: "nvidia/glm-4.7",
+    id: "free/glm-4.7",
     name: "[Free] GLM-4.7",
     version: "4.7",
     inputPrice: 0,
@@ -41273,7 +41191,7 @@ var BLOCKRUN_MODELS = [
     reasoning: true
   },
   {
-    id: "nvidia/llama-4-maverick",
+    id: "free/llama-4-maverick",
     name: "[Free] Llama 4 Maverick",
     version: "4-maverick",
     inputPrice: 0,
@@ -46849,8 +46767,6 @@ var BLOCKRUN_SOLANA_API = "https://sol.blockrun.ai/api";
 var IMAGE_DIR = join8(homedir5(), ".openclaw", "blockrun", "images");
 var AUTO_MODEL = "blockrun/auto";
 var ROUTING_PROFILES = /* @__PURE__ */ new Set([
-  "blockrun/free",
-  "free",
   "blockrun/eco",
   "eco",
   "blockrun/auto",
@@ -46858,39 +46774,26 @@ var ROUTING_PROFILES = /* @__PURE__ */ new Set([
   "blockrun/premium",
   "premium"
 ]);
-var FREE_MODEL = "nvidia/gpt-oss-120b";
+var FREE_MODEL = "free/gpt-oss-120b";
 var FREE_MODELS = /* @__PURE__ */ new Set([
-  "nvidia/gpt-oss-120b",
-  "nvidia/gpt-oss-20b",
-  "nvidia/nemotron-ultra-253b",
-  "nvidia/nemotron-3-super-120b",
-  "nvidia/nemotron-super-49b",
-  "nvidia/deepseek-v3.2",
-  "nvidia/mistral-large-3-675b",
-  "nvidia/qwen3-coder-480b",
-  "nvidia/devstral-2-123b",
-  "nvidia/glm-4.7",
-  "nvidia/llama-4-maverick"
+  "free/gpt-oss-120b",
+  "free/gpt-oss-20b",
+  "free/nemotron-ultra-253b",
+  "free/nemotron-3-super-120b",
+  "free/nemotron-super-49b",
+  "free/deepseek-v3.2",
+  "free/mistral-large-3-675b",
+  "free/qwen3-coder-480b",
+  "free/devstral-2-123b",
+  "free/glm-4.7",
+  "free/llama-4-maverick"
 ]);
-var FREE_TIER_CONFIGS = {
-  SIMPLE: {
-    primary: "nvidia/gpt-oss-20b",
-    fallback: ["nvidia/gpt-oss-120b", "nvidia/nemotron-super-49b"]
-  },
-  MEDIUM: {
-    primary: "nvidia/deepseek-v3.2",
-    fallback: ["nvidia/gpt-oss-120b", "nvidia/nemotron-super-49b"]
-  },
-  COMPLEX: {
-    primary: "nvidia/nemotron-ultra-253b",
-    fallback: ["nvidia/mistral-large-3-675b", "nvidia/deepseek-v3.2", "nvidia/gpt-oss-120b"]
-  },
-  REASONING: {
-    primary: "nvidia/nemotron-ultra-253b",
-    fallback: ["nvidia/nemotron-3-super-120b", "nvidia/deepseek-v3.2"]
+function toUpstreamModelId(modelId) {
+  if (modelId.startsWith("free/")) {
+    return "nvidia/" + modelId.slice("free/".length);
   }
-};
-var freeRequestCount = 0;
+  return modelId;
+}
 var MAX_MESSAGES = 200;
 var CONTEXT_LIMIT_KB = 5120;
 var HEARTBEAT_INTERVAL_MS = 2e3;
@@ -48252,7 +48155,7 @@ async function tryModelRequest(upstreamUrl, method, headers, body, modelId, maxT
   let requestBody = body;
   try {
     const parsed = JSON.parse(body.toString());
-    parsed.model = modelId;
+    parsed.model = toUpstreamModelId(modelId);
     if (Array.isArray(parsed.messages)) {
       parsed.messages = normalizeMessageRoles(parsed.messages);
     }
@@ -48396,7 +48299,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
         const estimatedTokens = Math.ceil(fullText.length / 4);
         const normalizedModel2 = typeof parsed.model === "string" ? parsed.model.trim().toLowerCase() : "";
         const profileName = normalizedModel2.replace("blockrun/", "");
-        const debugProfile = ["free", "eco", "auto", "premium"].includes(profileName) ? profileName : "auto";
+        const debugProfile = ["eco", "auto", "premium"].includes(profileName) ? profileName : "auto";
         const scoring = classifyByRules(
           debugPrompt,
           systemPrompt,
@@ -49016,20 +48919,15 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
             }
           }
           options.onRouted?.(routingDecision);
-          if (routingProfile === "free") {
-            freeRequestCount++;
-            if (freeRequestCount % 5 === 0) {
-              balanceFallbackNotice = `> **\u{1F4A1} Tip:** Free tier gives you 11 NVIDIA models. Want Claude, GPT-5, or Gemini? Fund your wallet \u2014 starting at $0.001/request.
-
-`;
-            }
-          }
         }
       }
       if (!effectiveSessionId && parsedMessages.length > 0) {
         effectiveSessionId = deriveSessionId(parsedMessages);
       }
       if (bodyModified) {
+        if (parsed.model && typeof parsed.model === "string") {
+          parsed.model = toUpstreamModelId(parsed.model);
+        }
         body = Buffer.from(JSON.stringify(parsed));
       }
     } catch (err) {
@@ -49128,28 +49026,19 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
       const sufficiency = await balanceMonitor.checkSufficient(bufferedCostMicros);
       if (sufficiency.info.isEmpty || !sufficiency.sufficient) {
         const originalModel = modelId;
-        const fallbackTier = routingDecision?.tier ?? "SIMPLE";
-        const freeTierConfig = FREE_TIER_CONFIGS[fallbackTier];
-        const freeModel = freeTierConfig.primary;
         console.log(
-          `[ClawRouter] Wallet ${sufficiency.info.isEmpty ? "empty" : "insufficient"} (${sufficiency.info.balanceUSD}), falling back to free model: ${freeModel} (tier: ${fallbackTier}, requested: ${originalModel})`
+          `[ClawRouter] Wallet ${sufficiency.info.isEmpty ? "empty" : "insufficient"} (${sufficiency.info.balanceUSD}), falling back to free model: ${FREE_MODEL} (requested: ${originalModel})`
         );
-        modelId = freeModel;
+        modelId = FREE_MODEL;
         isFreeModel = true;
         const parsed = JSON.parse(body.toString());
-        parsed.model = freeModel;
+        parsed.model = toUpstreamModelId(FREE_MODEL);
         body = Buffer.from(JSON.stringify(parsed));
         balanceFallbackNotice = sufficiency.info.isEmpty ? `> **\u26A0\uFE0F Wallet empty** \u2014 using free model. Fund your wallet to use ${originalModel}.
 
 ` : `> **\u26A0\uFE0F Insufficient balance** (${sufficiency.info.balanceUSD}) \u2014 using free model instead of ${originalModel}.
 
 `;
-        freeRequestCount++;
-        if (freeRequestCount % 5 === 0) {
-          balanceFallbackNotice = `> **\u{1F4A1} Tip:** Free tier gives you 11 NVIDIA models. Want Claude, GPT-5, or Gemini? Fund your wallet \u2014 starting at $0.001/request.
-
-`;
-        }
         options.onLowBalance?.({
           balanceUSD: sufficiency.info.balanceUSD,
           walletAddress: sufficiency.info.walletAddress