🤖 chore: update models.json from LiteLLM

bun.lock

@@ -1,16 +1,15 @@
 {
   "lockfileVersion": 1,
-  "configVersion": 0,
   "workspaces": {
     "": {
       "name": "@coder/cmux",
       "dependencies": {
-        "@ai-sdk/anthropic": "^2.0.44",
-        "@ai-sdk/google": "^2.0.38",
-        "@ai-sdk/openai": "^2.0.66",
-        "@ai-sdk/xai": "^2.0.33",
+        "@ai-sdk/anthropic": "^2.0.47",
+        "@ai-sdk/google": "^2.0.43",
+        "@ai-sdk/openai": "^2.0.72",
+        "@ai-sdk/xai": "^2.0.36",
         "@lydell/node-pty": "1.1.0",
-        "@openrouter/ai-sdk-provider": "^1.2.2",
+        "@openrouter/ai-sdk-provider": "^1.2.5",
         "@radix-ui/react-checkbox": "^1.3.3",
         "@radix-ui/react-dialog": "^1.1.15",
         "@radix-ui/react-dropdown-menu": "^2.1.16",
@@ -22,7 +21,7 @@
         "@radix-ui/react-tabs": "^1.1.13",
         "@radix-ui/react-toggle-group": "^1.1.11",
         "@radix-ui/react-tooltip": "^1.2.8",
-        "ai": "^5.0.93",
+        "ai": "^5.0.101",
         "ai-tokenizer": "^1.0.4",
         "chalk": "^5.6.2",
         "cors": "^2.8.5",
@@ -137,21 +136,21 @@
 
     "@adobe/css-tools": ["@adobe/css-tools@4.4.4", "", {}, "sha512-Elp+iwUx5rN5+Y8xLt5/GRoG20WGoDCQ/1Fb+1LiGtvwbDavuSk0jhD/eZdckHAuzcDzccnkv+rEjyWfRx18gg=="],
 
-    "@ai-sdk/anthropic": ["@ai-sdk/anthropic@2.0.44", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-o8TfNXRzO/KZkBrcx+CL9LQsPhx7PHyqzUGjza3TJaF9WxfH1S5UQLAmEw8F7lQoHNLU0IX03WT8o8R/4JbUxQ=="],
+    "@ai-sdk/anthropic": ["@ai-sdk/anthropic@2.0.47", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-YioBDTTQ6z2fijcOByG6Gj7me0ITqaJACprHROis7fXFzYIBzyAwxhsCnOrXO+oXv+9Ixddgy/Cahdmu84uRvQ=="],
 
-    "@ai-sdk/gateway": ["@ai-sdk/gateway@2.0.10", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17", "@vercel/oidc": "3.0.3" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-c++qOKfjKokTPAJ+vP9UXXNuTQ819yEDCZVXBhpZbgRly1P4fHTJbIAwuh+Qxxe9Bmtu8PEta0JGYZxc+hm7/Q=="],
+    "@ai-sdk/gateway": ["@ai-sdk/gateway@2.0.15", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17", "@vercel/oidc": "3.0.5" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-i1YVKzC1dg9LGvt+GthhD7NlRhz9J4+ZRj3KELU14IZ/MHPsOBiFeEoCCIDLR+3tqT8/+5nIsK3eZ7DFRfMfdw=="],
 
-    "@ai-sdk/google": ["@ai-sdk/google@2.0.38", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-z+RFCxRA/dSd3eCkGBlnk79nz3jv8vwaW42gVc+qDuMofNfvjRz19rjnkFNuYQ6cEUcPKCo0P1rD/JLeTN2Z5A=="],
+    "@ai-sdk/google": ["@ai-sdk/google@2.0.43", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-qO6giuoYCX/SdZScP/3VO5Xnbd392zm3HrTkhab/efocZU8J/VVEAcAUE1KJh0qOIAYllofRtpJIUGkRK8Q5rw=="],
 
-    "@ai-sdk/openai": ["@ai-sdk/openai@2.0.68", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-qUSLFkqgUoFArzBwttu0KWVAZYjbsdZGOklSJXpfZ2nDC61yseHxtcnuG8u6tqKnGXDh4eakEgREDWU2sRht7A=="],
+    "@ai-sdk/openai": ["@ai-sdk/openai@2.0.72", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-9j8Gdt9gFiUGFdQIjjynbC7+w8YQxkXje6dwAq1v2Pj17wmB3U0Td3lnEe/a+EnEysY3mdkc8dHPYc5BNev9NQ=="],
 
     "@ai-sdk/openai-compatible": ["@ai-sdk/openai-compatible@1.0.27", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-bpYruxVLhrTbVH6CCq48zMJNeHu6FmHtEedl9FXckEgcIEAi036idFhJlcRwC1jNCwlacbzb8dPD7OAH1EKJaQ=="],
 
     "@ai-sdk/provider": ["@ai-sdk/provider@2.0.0", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-6o7Y2SeO9vFKB8lArHXehNuusnpddKPk7xqL7T2/b+OvXMRIXUO1rR4wcv1hAFUAT9avGZshty3Wlua/XA7TvA=="],
 
     "@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@3.0.17", "", { "dependencies": { "@ai-sdk/provider": "2.0.0", "@standard-schema/spec": "^1.0.0", "eventsource-parser": "^3.0.6" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-TR3Gs4I3Tym4Ll+EPdzRdvo/rc8Js6c4nVhFLuvGLX/Y4V9ZcQMa/HTiYsHEgmYrf1zVi6Q145UEZUfleOwOjw=="],
 
-    "@ai-sdk/xai": ["@ai-sdk/xai@2.0.33", "", { "dependencies": { "@ai-sdk/openai-compatible": "1.0.27", "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-0+S+hxbAj8dA8/3dYQsmgkVkPcs8yptO1ueLWtJpa6PYjrdyliDcPSCZREL8aE76vHGvFsYlRABFfH9Ps2M8tg=="],
+    "@ai-sdk/xai": ["@ai-sdk/xai@2.0.36", "", { "dependencies": { "@ai-sdk/openai-compatible": "1.0.27", "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-tQuCDVNK4W4fiom59r2UnU7u9SAz58fpl5yKYoS9IbMOrDRO3fzQGWmj2p8MUvz9LzXf6hiyUkVNFGzzx+uZcw=="],
 
     "@antfu/install-pkg": ["@antfu/install-pkg@1.1.0", "", { "dependencies": { "package-manager-detector": "^1.3.0", "tinyexec": "^1.0.1" } }, "sha512-MGQsmw10ZyI+EJo45CdSER4zEb+p31LpDAFp2Z3gkSd1yqVZGi0Ebx++YTEMonJy4oChEMLsxZ64j8FH6sSqtQ=="],
 
@@ -523,7 +522,7 @@
 
     "@npmcli/move-file": ["@npmcli/move-file@2.0.1", "", { "dependencies": { "mkdirp": "^1.0.4", "rimraf": "^3.0.2" } }, "sha512-mJd2Z5TjYWq/ttPLLGqArdtnC74J6bOzg4rMDnN+p1xTacZ2yPRCk2y0oSWQtygLR9YVQXgOcONrwtnk3JupxQ=="],
 
-    "@openrouter/ai-sdk-provider": ["@openrouter/ai-sdk-provider@1.2.3", "", { "dependencies": { "@openrouter/sdk": "^0.1.8" }, "peerDependencies": { "ai": "^5.0.0", "zod": "^3.24.1 || ^v4" } }, "sha512-a6Nc8dPRHakRH9966YJ/HZJhLOds7DuPTscNZDoAr+Aw+tEFUlacSJMvb/b3gukn74mgbuaJRji9YOn62ipfVg=="],
+    "@openrouter/ai-sdk-provider": ["@openrouter/ai-sdk-provider@1.2.5", "", { "dependencies": { "@openrouter/sdk": "^0.1.8" }, "peerDependencies": { "ai": "^5.0.0", "zod": "^3.24.1 || ^v4" } }, "sha512-NrvJFPvdEUo6DYUQIVWPGfhafuZ2PAIX7+CUMKGknv8TcTNVo0TyP1y5SU7Bgjf/Wup9/74UFKUB07icOhVZjQ=="],
 
     "@openrouter/sdk": ["@openrouter/sdk@0.1.11", "", { "dependencies": { "zod": "^3.25.0 || ^4.0.0" }, "peerDependencies": { "@tanstack/react-query": "^5", "react": "^18 || ^19", "react-dom": "^18 || ^19" }, "optionalPeers": ["@tanstack/react-query", "react", "react-dom"] }, "sha512-OuPc8qqidL/PUM8+9WgrOfSR9+b6rKIWiezGcUJ54iPTdh+Gye5Qjut6hrLWlOCMZE7Z853gN90r1ft4iChj7Q=="],
 
@@ -1085,7 +1084,7 @@
 
     "@unrs/resolver-binding-win32-x64-msvc": ["@unrs/resolver-binding-win32-x64-msvc@1.11.1", "", { "os": "win32", "cpu": "x64" }, "sha512-lrW200hZdbfRtztbygyaq/6jP6AKE8qQN2KvPcJ+x7wiD038YtnYtZ82IMNJ69GJibV7bwL3y9FgK+5w/pYt6g=="],
 
-    "@vercel/oidc": ["@vercel/oidc@3.0.3", "", {}, "sha512-yNEQvPcVrK9sIe637+I0jD6leluPxzwJKx/Haw6F4H77CdDsszUn5V3o96LPziXkSNE2B83+Z3mjqGKBK/R6Gg=="],
+    "@vercel/oidc": ["@vercel/oidc@3.0.5", "", {}, "sha512-fnYhv671l+eTTp48gB4zEsTW/YtRgRPnkI2nT7x6qw5rkI1Lq2hTmQIpHPgyThI0znLK+vX2n9XxKdXZ7BUbbw=="],
 
     "@vitejs/plugin-react": ["@vitejs/plugin-react@4.7.0", "", { "dependencies": { "@babel/core": "^7.28.0", "@babel/plugin-transform-react-jsx-self": "^7.27.1", "@babel/plugin-transform-react-jsx-source": "^7.27.1", "@rolldown/pluginutils": "1.0.0-beta.27", "@types/babel__core": "^7.20.5", "react-refresh": "^0.17.0" }, "peerDependencies": { "vite": "^4.2.0 || ^5.0.0 || ^6.0.0 || ^7.0.0" } }, "sha512-gUu9hwfWvvEDBBmgtAowQCojwZmJ5mcLn3aufeCsitijs3+f2NsrPtlAWIR6OPiqljl96GVCUbLe0HyqIpVaoA=="],
 
@@ -1115,7 +1114,7 @@
 
     "aggregate-error": ["aggregate-error@3.1.0", "", { "dependencies": { "clean-stack": "^2.0.0", "indent-string": "^4.0.0" } }, "sha512-4I7Td01quW/RpocfNayFdFVk1qSuoh0E7JrbRJ16nH01HhKFQ88INq9Sd+nd72zqRySlr9BmDA8xlEJ6vJMrYA=="],
 
-    "ai": ["ai@5.0.94", "", { "dependencies": { "@ai-sdk/gateway": "2.0.10", "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17", "@opentelemetry/api": "1.9.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-PQs3H8xDhORG/L6eyR+SxAKUsa0ORO4ENvRovzGgPmPGCIlwle6UD5VIMQFtj1gvZk+BKBUVEFFwtkTeJTAURw=="],
+    "ai": ["ai@5.0.101", "", { "dependencies": { "@ai-sdk/gateway": "2.0.15", "@ai-sdk/provider": "2.0.0", "@ai-sdk/provider-utils": "3.0.17", "@opentelemetry/api": "1.9.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-/P4fgs2PGYTBaZi192YkPikOudsl9vccA65F7J7LvoNTOoP5kh1yAsJPsKAy6FXU32bAngai7ft1UDyC3u7z5g=="],
 
     "ai-tokenizer": ["ai-tokenizer@1.0.4", "", { "peerDependencies": { "ai": "^5.0.0" }, "optionalPeers": ["ai"] }, "sha512-BHOUljsmH0SEO9bULQL3sz6pJ4jv00r+NHxX3kR6tn1suAAj6DDN4njSk+sqCOI5Cm6FqizUhDfoYZ0R+5/WVQ=="],
 

package.json

@@ -45,12 +45,12 @@
     "postinstall": "sh scripts/postinstall.sh"
   },
   "dependencies": {
-    "@ai-sdk/anthropic": "^2.0.44",
-    "@ai-sdk/google": "^2.0.38",
-    "@ai-sdk/openai": "^2.0.66",
-    "@ai-sdk/xai": "^2.0.33",
+    "@ai-sdk/anthropic": "^2.0.47",
+    "@ai-sdk/google": "^2.0.43",
+    "@ai-sdk/openai": "^2.0.72",
+    "@ai-sdk/xai": "^2.0.36",
     "@lydell/node-pty": "1.1.0",
-    "@openrouter/ai-sdk-provider": "^1.2.2",
+    "@openrouter/ai-sdk-provider": "^1.2.5",
     "@radix-ui/react-checkbox": "^1.3.3",
     "@radix-ui/react-dialog": "^1.1.15",
     "@radix-ui/react-dropdown-menu": "^2.1.16",
@@ -62,7 +62,7 @@
     "@radix-ui/react-tabs": "^1.1.13",
     "@radix-ui/react-toggle-group": "^1.1.11",
     "@radix-ui/react-tooltip": "^1.2.8",
-    "ai": "^5.0.93",
+    "ai": "^5.0.101",
     "ai-tokenizer": "^1.0.4",
     "chalk": "^5.6.2",
     "cors": "^2.8.5",

src/common/utils/ai/providerOptions.ts

@@ -19,6 +19,7 @@ import {
 import { log } from "@/node/services/log";
 import type { MuxMessage } from "@/common/types/message";
 import { enforceThinkingPolicy } from "@/browser/utils/thinking/policy";
+import { getModelStats } from "@/common/utils/tokens/modelStats";
 
 /**
  * OpenRouter reasoning options
@@ -277,3 +278,77 @@ export function buildProviderOptions(
   log.debug("buildProviderOptions: Unsupported provider", provider);
   return {};
 }
+
+/**
+ * Calculate the effective maxOutputTokens for a model based on its limits and thinking budget
+ *
+ * For Anthropic models with extended thinking, the AI SDK adds thinkingBudget to maxOutputTokens
+ * internally. We need to ensure the sum doesn't exceed the model's max_output_tokens limit.
+ *
+ * For example, Claude Opus 4 has max_output_tokens=32000. If we use:
+ * - thinkingBudget=20000 (high)
+ * - maxOutputTokens=32000
+ * Then total=52000 which exceeds 32000 → SDK shows warning and caps output
+ *
+ * Solution: Reduce maxOutputTokens so that maxOutputTokens + thinkingBudget <= model limit
+ *
+ * @param modelString - Full model string (e.g., "anthropic:claude-opus-4-1")
+ * @param thinkingLevel - Current thinking level
+ * @param requestedMaxOutputTokens - Optional user-requested maxOutputTokens
+ * @returns Effective maxOutputTokens that respects model limits with thinking budget
+ */
+export function calculateEffectiveMaxOutputTokens(
+  modelString: string,
+  thinkingLevel: ThinkingLevel,
+  requestedMaxOutputTokens?: number
+): number | undefined {
+  const [provider] = modelString.split(":");
+
+  // Only apply this adjustment for Anthropic models
+  if (provider !== "anthropic") {
+    return requestedMaxOutputTokens;
+  }
+
+  // Get the actual thinking level after policy enforcement
+  const effectiveThinking = enforceThinkingPolicy(modelString, thinkingLevel);
+  const thinkingBudget = ANTHROPIC_THINKING_BUDGETS[effectiveThinking];
+
+  // Get model's max output tokens from models.json
+  const modelStats = getModelStats(modelString);
+  const modelMaxOutput = modelStats?.max_output_tokens;
+
+  // If we don't know the model's max output, return requested value
+  if (!modelMaxOutput) {
+    log.debug("calculateEffectiveMaxOutputTokens: Unknown model max output, using requested", {
+      modelString,
+      requestedMaxOutputTokens,
+    });
+    return requestedMaxOutputTokens;
+  }
+
+  // Calculate the maximum safe maxOutputTokens
+  // The SDK will add thinkingBudget to maxOutputTokens, so we need room for both
+  const maxSafeOutput = modelMaxOutput - thinkingBudget;
+
+  // If user didn't request specific tokens, use the max safe value
+  const targetOutput = requestedMaxOutputTokens ?? modelMaxOutput;
+
+  // Cap at the safe maximum
+  const effectiveOutput = Math.min(targetOutput, maxSafeOutput);
+
+  // Ensure we don't go below a reasonable minimum (1000 tokens)
+  const finalOutput = Math.max(effectiveOutput, 1000);
+
+  log.debug("calculateEffectiveMaxOutputTokens", {
+    modelString,
+    thinkingLevel,
+    effectiveThinking,
+    thinkingBudget,
+    modelMaxOutput,
+    requestedMaxOutputTokens,
+    maxSafeOutput,
+    finalOutput,
+  });
+
+  return finalOutput;
+}

src/common/utils/tokens/modelStats.ts

@@ -3,6 +3,7 @@ import { modelsExtra } from "./models-extra";
 
 export interface ModelStats {
   max_input_tokens: number;
+  max_output_tokens?: number;
   input_cost_per_token: number;
   output_cost_per_token: number;
   cache_creation_input_token_cost?: number;
@@ -11,6 +12,7 @@ export interface ModelStats {
 
 interface RawModelData {
   max_input_tokens?: number | string;
+  max_output_tokens?: number | string;
   input_cost_per_token?: number;
   output_cost_per_token?: number;
   cache_creation_input_token_cost?: number;
@@ -37,6 +39,8 @@ function extractModelStats(data: RawModelData): ModelStats {
   /* eslint-disable @typescript-eslint/non-nullable-type-assertion-style */
   return {
     max_input_tokens: data.max_input_tokens as number,
+    max_output_tokens:
+      typeof data.max_output_tokens === "number" ? data.max_output_tokens : undefined,
     input_cost_per_token: data.input_cost_per_token as number,
     output_cost_per_token: data.output_cost_per_token as number,
     cache_creation_input_token_cost:

src/common/utils/tokens/models-extra.ts

@@ -22,15 +22,16 @@ interface ModelData {
 }
 
 export const modelsExtra: Record<string, ModelData> = {
-  // Claude Opus 4.5 - Released November 2025
-  // $15/M input, $75/M output (same pricing as Opus 4.1)
+  // Claude Opus 4.5 - Released November 24, 2025
+  // $5/M input, $25/M output (price drop from Opus 4.1's $15/$75)
+  // 64K max output tokens (matches Sonnet 4.5)
   "claude-opus-4-5": {
     max_input_tokens: 200000,
-    max_output_tokens: 32000,
-    input_cost_per_token: 0.000015, // $15 per million input tokens
-    output_cost_per_token: 0.000075, // $75 per million output tokens
-    cache_creation_input_token_cost: 0.00001875, // $18.75 per million tokens
-    cache_read_input_token_cost: 0.0000015, // $1.50 per million tokens
+    max_output_tokens: 64000,
+    input_cost_per_token: 0.000005, // $5 per million input tokens
+    output_cost_per_token: 0.000025, // $25 per million output tokens
+    cache_creation_input_token_cost: 0.00000625, // $6.25 per million tokens (estimated)
+    cache_read_input_token_cost: 0.0000005, // $0.50 per million tokens (estimated)
     litellm_provider: "anthropic",
     mode: "chat",
     supports_function_calling: true,