feat: add 'shape' field to provider so models can specify if they use responses vs completions apis (use only if model only supports 1 of)

sync nano-gpt models with live API catalog

.github/workflows/opencode.yml

@@ -7,8 +7,10 @@ on:
 jobs:
   opencode:
     if: |
-      contains(github.event.comment.body, '/oc') ||
-      contains(github.event.comment.body, '/opencode')
+      contains(github.event.comment.body, ' /oc') ||
+      startsWith(github.event.comment.body, '/oc') ||
+      contains(github.event.comment.body, ' /opencode') ||
+      startsWith(github.event.comment.body, '/opencode')
     runs-on: ubuntu-latest
     permissions:
       contents: read
@@ -22,4 +24,4 @@ jobs:
         env:
           ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
         with:
-          model: anthropic/claude-sonnet-4-20250514
+          model: anthropic/claude-sonnet-4-20250514

.gitignore

@@ -1,5 +1,9 @@
 .env
 .sst
+.idea
 dist
 .DS_Store
 node_modules
+data/tokenspeed-monitor.sqlite
+data/tokenspeed-monitor.sqlite-shm
+data/tokenspeed-monitor.sqlite-wal

package.json

@@ -17,7 +17,8 @@
   "scripts": {
     "validate": "bun ./packages/core/script/validate.ts",
     "helicone:generate": "bun ./packages/core/script/generate-helicone.ts",
-    "venice:generate": "bun ./packages/core/script/generate-venice.ts"
+    "venice:generate": "bun ./packages/core/script/generate-venice.ts",
+    "vercel:generate": "bun ./packages/core/script/generate-vercel.ts"
   },
   "dependencies": {
     "@cloudflare/workers-types": "^4.20250801.0",

packages/core/script/generate-friendli.ts

@@ -48,8 +48,8 @@ const familyPatterns: [RegExp, string][] = [
   [/llama-4/i, "llama-4"],
   [/qwen3/i, "qwen3"],
   [/deepseek-r1/i, "deepseek-r1"],
-  [/exaone/i, "exaone"],
   [/glm-4/i, "glm-4"],
+  [/glm-5/i, "glm"],
 ];
 
 function inferFamily(modelId: string, modelName: string): string | undefined {

packages/core/script/generate-ollama-cloud.ts

@@ -35,6 +35,31 @@ type OllamaModel = Omit<Model, "id"> & {
   limit: Model["limit"] & { output?: number };
 };
 
+type ComparableModel = Pick<Model,
+  | "name"
+  | "attachment"
+  | "reasoning"
+  | "tool_call"
+  | "knowledge"
+  | "open_weights"
+  | "modalities"
+> & {
+  limit: Pick<Model["limit"], "context">;
+};
+
+function normalizeForComparison(model: Omit<Model, "id">): ComparableModel {
+  return {
+    name: model.name,
+    attachment: model.attachment,
+    reasoning: model.reasoning,
+    tool_call: model.tool_call,
+    knowledge: model.knowledge,
+    open_weights: model.open_weights,
+    limit: { context: model.limit.context },
+    modalities: model.modalities,
+  };
+}
+
 const OllamaTagsResponse = z.object({
   models: z.array(
     z.object({
@@ -137,20 +162,34 @@ for (const modelName of modelNames) {
   modelsData.push({ name: modelName, data: showParsed.data });
 }
 
-console.log(`Fetched all models. Writing new files...`);
+console.log(`Fetched all models. Syncing files...`);
+
+const existingFiles = Array.from(new Bun.Glob("*.toml").scanSync(modelsDir));
+const existingModelNames = new Set(existingFiles.map((f) => f.replace(/\.toml$/, "")));
+const apiModelNames = new Set(modelNames);
+
+let deleted = 0;
+for (const existingName of existingModelNames) {
+  if (!apiModelNames.has(existingName)) {
+    const filePath = path.join(modelsDir, modelFileName(existingName));
+    await Bun.file(filePath).delete();
+    console.log(`Deleted: ${modelFileName(existingName)}`);
+    deleted++;
+  }
+}
 
 let created = 0;
+let skipped = 0;
 for (const { name, data } of modelsData) {
   const fileName = modelFileName(name);
   const filePath = path.join(modelsDir, fileName);
 
-  let existingData: Omit<Model, "id"> | null;
+  let existingData: Omit<Model, "id"> | null = null;
   try {
     const existingToml = await Bun.file(filePath).text();
     existingData = Bun.TOML.parse(existingToml) as Omit<Model, "id">;
   } catch {
     // File doesn't exist
-    existingData = null;
   }
 
   const family = existingData?.family ?? (data.details.family as ModelFamily);
@@ -179,9 +218,20 @@ for (const { name, data } of modelsData) {
     },
   };
 
+  if (existingData) {
+    const normalizedExisting = normalizeForComparison(existingData);
+    const normalizedIncoming = normalizeForComparison(ollamaModel);
+
+    if (Bun.deepEquals(normalizedExisting, normalizedIncoming)) {
+      console.log(`Skipped (no changes): ${fileName}`);
+      skipped++;
+      continue;
+    }
+  }
+
   await Bun.write(filePath, generateToml(name, ollamaModel));
   console.log(`Created: ${fileName}`);
   created++;
 }
 
-console.log(`\nDone. Created ${created}`);
+console.log(`\nDone. Created: ${created}, Skipped: ${skipped}, Deleted: ${deleted}`);

packages/core/script/generate-venice.ts

@@ -3,30 +3,11 @@
 import { z } from "zod";
 import path from "node:path";
 import { readdir } from "node:fs/promises";
-import * as readline from "node:readline";
 import { ModelFamilyValues } from "../src/family.js";
 
 // Venice API endpoint
 const API_ENDPOINT = "https://api.venice.ai/api/v1/models?type=text";
 
-async function promptForApiKey(): Promise<string | null> {
-  const rl = readline.createInterface({
-    input: process.stdin,
-    output: process.stdout,
-  });
-
-  return new Promise((resolve) => {
-    rl.question(
-      "Enter Venice API key to include alpha models (or press Enter to skip): ",
-      (answer) => {
-        rl.close();
-        const trimmed = answer.trim();
-        resolve(trimmed.length > 0 ? trimmed : null);
-      },
-    );
-  });
-}
-
 // Zod schemas for API response validation
 const Capabilities = z
   .object({
@@ -43,12 +24,25 @@ const Capabilities = z
   })
   .passthrough();
 
+const PricingTier = z.object({ usd: z.number(), diem: z.number().optional() }).passthrough();
+
+const ExtendedPricing = z
+  .object({
+    context_token_threshold: z.number(),
+    input: PricingTier,
+    output: PricingTier,
+    cache_input: PricingTier.optional(),
+    cache_write: PricingTier.optional(),
+  })
+  .passthrough();
+
 const Pricing = z
   .object({
-    input: z.object({ usd: z.number(), diem: z.number().optional() }).passthrough(),
-    output: z.object({ usd: z.number(), diem: z.number().optional() }).passthrough(),
-    cache_input: z.object({ usd: z.number(), diem: z.number().optional() }).passthrough().optional(),
-    cache_write: z.object({ usd: z.number(), diem: z.number().optional() }).passthrough().optional(),
+    input: PricingTier,
+    output: PricingTier,
+    cache_input: PricingTier.optional(),
+    cache_write: PricingTier.optional(),
+    extended: ExtendedPricing.optional(),
   })
   .passthrough();
 
@@ -56,6 +50,7 @@ const ModelSpec = z
   .object({
     pricing: Pricing.optional(),
     availableContextTokens: z.number(),
+    maxCompletionTokens: z.number().optional(),
     capabilities: Capabilities,
     constraints: z.any().optional(),
     name: z.string(),
@@ -162,6 +157,12 @@ interface ExistingModel {
     reasoning?: number;
     cache_read?: number;
     cache_write?: number;
+    context_over_200k?: {
+      input?: number;
+      output?: number;
+      cache_read?: number;
+      cache_write?: number;
+    };
   };
   limit?: {
     context?: number;
@@ -213,6 +214,12 @@ interface MergedModel {
     output: number;
     cache_read?: number;
     cache_write?: number;
+    context_over_200k?: {
+      input: number;
+      output: number;
+      cache_read?: number;
+      cache_write?: number;
+    };
   };
   limit: {
     context: number;
@@ -232,11 +239,7 @@ function mergeModel(
   const caps = spec.capabilities;
 
   const contextTokens = spec.availableContextTokens;
-  const proposedOutputTokens = Math.floor(contextTokens / 4);
-  const outputTokens =
-    existing?.limit?.output !== undefined && existing.limit.output < proposedOutputTokens
-      ? existing.limit.output
-      : proposedOutputTokens
+  const outputTokens = spec.maxCompletionTokens ?? Math.floor(contextTokens / 4);
 
   const openWeights = spec.modelSource
     ? spec.modelSource.toLowerCase().includes("huggingface")
@@ -285,6 +288,16 @@ function mergeModel(
       ...(spec.pricing.cache_input && { cache_read: spec.pricing.cache_input.usd }),
       ...(spec.pricing.cache_write && { cache_write: spec.pricing.cache_write.usd }),
     };
+
+    // Extended pricing maps to context_over_200k
+    if (spec.pricing.extended) {
+      merged.cost.context_over_200k = {
+        input: spec.pricing.extended.input.usd,
+        output: spec.pricing.extended.output.usd,
+        ...(spec.pricing.extended.cache_input && { cache_read: spec.pricing.extended.cache_input.usd }),
+        ...(spec.pricing.extended.cache_write && { cache_write: spec.pricing.extended.cache_write.usd }),
+      };
+    }
   }
 
   const inferred = inferFamily(apiModel.id, spec.name);
@@ -352,6 +365,19 @@ function formatToml(model: MergedModel): string {
     if (model.cost.cache_write !== undefined) {
       lines.push(`cache_write = ${model.cost.cache_write}`);
     }
+
+    if (model.cost.context_over_200k) {
+      lines.push("");
+      lines.push(`[cost.context_over_200k]`);
+      lines.push(`input = ${model.cost.context_over_200k.input}`);
+      lines.push(`output = ${model.cost.context_over_200k.output}`);
+      if (model.cost.context_over_200k.cache_read !== undefined) {
+        lines.push(`cache_read = ${model.cost.context_over_200k.cache_read}`);
+      }
+      if (model.cost.context_over_200k.cache_write !== undefined) {
+        lines.push(`cache_write = ${model.cost.context_over_200k.cache_write}`);
+      }
+    }
   }
 
   // Limit section
@@ -414,6 +440,10 @@ function detectChanges(
   compare("cost.output", existing.cost?.output, merged.cost?.output);
   compare("cost.cache_read", existing.cost?.cache_read, merged.cost?.cache_read);
   compare("cost.cache_write", existing.cost?.cache_write, merged.cost?.cache_write);
+  compare("cost.context_over_200k.input", existing.cost?.context_over_200k?.input, merged.cost?.context_over_200k?.input);
+  compare("cost.context_over_200k.output", existing.cost?.context_over_200k?.output, merged.cost?.context_over_200k?.output);
+  compare("cost.context_over_200k.cache_read", existing.cost?.context_over_200k?.cache_read, merged.cost?.context_over_200k?.cache_read);
+  compare("cost.context_over_200k.cache_write", existing.cost?.context_over_200k?.cache_write, merged.cost?.context_over_200k?.cache_write);
   compare("limit.context", existing.limit?.context, merged.limit.context);
   compare("limit.output", existing.limit?.output, merged.limit.output);
   compare("modalities.input", existing.modalities?.input, merged.modalities.input);
@@ -435,7 +465,7 @@ async function main() {
     "models",
   );
 
-  // Check for API key from CLI argument, environment, or prompt
+  // Check for API key from CLI argument or environment variable
   let apiKey: string | null = null;
 
   // Check CLI args for --api-key=xxx or --api-key xxx
@@ -454,11 +484,6 @@ async function main() {
     apiKey = process.env.VENICE_API_KEY ?? null;
   }
 
-  // Prompt if still no key
-  if (!apiKey) {
-    apiKey = await promptForApiKey();
-  }
-
   const includeAlpha = apiKey !== null;
 
   if (dryRun) {

packages/core/script/generate-vercel.ts

@@ -0,0 +1,572 @@
+#!/usr/bin/env bun
+
+/**
+ * Generates Vercel model TOML files from the AI Gateway API.
+ *
+ * Flags:
+ * --dry-run: Preview changes without writing files
+ * --new-only: Only create new models, skip updating existing ones
+ */
+
+import { z } from "zod";
+import path from "node:path";
+import { mkdir } from "node:fs/promises";
+import { ModelFamilyValues } from "../src/family.js";
+
+const API_ENDPOINT = "https://ai-gateway.vercel.sh/v1/models";
+
+enum ModelType {
+  Language = "language",
+  Embedding = "embedding",
+  Image = "image",
+  Video = "video",
+}
+
+enum SkipZeroFields {
+  LimitContext = "limit.context",
+  LimitOutput = "limit.output",
+}
+
+const PricingTier = z.object({
+  cost: z.string(),
+  min: z.number(),
+  max: z.number().optional(),
+});
+
+const Pricing = z.object({
+  input: z.string().optional(),
+  output: z.string().optional(),
+  input_cache_read: z.string().optional(),
+  input_cache_write: z.string().optional(),
+  input_tiers: z.array(PricingTier).optional(),
+  output_tiers: z.array(PricingTier).optional(),
+  input_cache_read_tiers: z.array(PricingTier).optional(),
+  input_cache_write_tiers: z.array(PricingTier).optional(),
+}).passthrough();
+
+const VercelModel = z.object({
+  id: z.string(),
+  name: z.string(),
+  created: z.number(),
+  released: z.number().optional(),
+  context_window: z.number(),
+  max_tokens: z.number(),
+  type: z.nativeEnum(ModelType),
+  tags: z.array(z.string()).optional().default([]),
+  pricing: Pricing.optional(),
+}).passthrough();
+
+const VercelResponse = z.object({
+  data: z.array(VercelModel),
+}).passthrough();
+
+interface ExistingModel {
+  name?: string;
+  family?: string;
+  attachment?: boolean;
+  reasoning?: boolean;
+  tool_call?: boolean;
+  structured_output?: boolean;
+  temperature?: boolean;
+  knowledge?: string;
+  release_date?: string;
+  last_updated?: string;
+  open_weights?: boolean;
+  interleaved?: boolean | { field: string };
+  status?: string;
+  cost?: {
+    input?: number;
+    output?: number;
+    cache_read?: number;
+    cache_write?: number;
+  };
+  limit?: {
+    context?: number;
+    output?: number;
+  };
+  modalities?: {
+    input?: string[];
+    output?: string[];
+  };
+}
+
+interface MergedModel {
+  name: string;
+  family?: string;
+  attachment: boolean;
+  reasoning: boolean;
+  tool_call: boolean;
+  structured_output?: boolean;
+  temperature: boolean;
+  knowledge?: string;
+  release_date: string;
+  last_updated: string;
+  open_weights: boolean;
+  interleaved?: boolean | { field: string };
+  status?: string;
+  cost?: {
+    input: number;
+    output: number;
+    cache_read?: number;
+    cache_write?: number;
+  };
+  limit: {
+    context: number;
+    output: number;
+  };
+  modalities: {
+    input: string[];
+    output: string[];
+  };
+}
+
+interface Changes {
+  field: string;
+  oldValue: string;
+  newValue: string;
+}
+
+function timestampToDate(timestamp: number): string {
+  const date = new Date(timestamp * 1000);
+  return date.toISOString().slice(0, 10);
+}
+
+function getTodayDate(): string {
+  return new Date().toISOString().slice(0, 10);
+}
+
+// Number utilities
+function formatNumber(n: number): string {
+  if (n >= 1000) {
+    return n.toString().replace(/\B(?=(\d{3})+(?!\d))/g, "_");
+  }
+  return n.toString();
+}
+
+function isSubstring(target: string, family: string): boolean {
+  return target.toLowerCase().includes(family.toLowerCase());
+}
+
+function matchesFamily(target: string, family: string): boolean {
+  const targetLower = target.toLowerCase();
+  const familyLower = family.toLowerCase();
+  let familyIdx = 0;
+
+  for (let i = 0; i < targetLower.length && familyIdx < familyLower.length; i++) {
+    if (targetLower[i] === familyLower[familyIdx]) {
+      familyIdx++;
+    }
+  }
+
+  return familyIdx === familyLower.length;
+}
+
+function inferFamily(modelId: string, modelName: string): string | undefined {
+  const sortedFamilies = [...ModelFamilyValues].sort((a, b) => b.length - a.length);
+
+  // First pass: try exact substring matches
+  for (const family of sortedFamilies) {
+    if (isSubstring(modelId, family)) {
+      return family;
+    }
+  }
+
+  for (const family of sortedFamilies) {
+    if (isSubstring(modelName, family)) {
+      return family;
+    }
+  }
+
+  // Second pass: fall back to subsequence matching
+  for (const family of sortedFamilies) {
+    if (matchesFamily(modelId, family)) {
+      return family;
+    }
+  }
+
+  for (const family of sortedFamilies) {
+    if (matchesFamily(modelName, family)) {
+      return family;
+    }
+  }
+
+  return undefined;
+}
+
+function buildInputModalities(tags: string[]): string[] {
+  const mods: string[] = ["text"];
+  const tagSet = new Set(tags);
+
+  if (tagSet.has("vision")) mods.push("image");
+  if (tagSet.has("file-input")) mods.push("pdf");
+
+  return mods;
+}
+
+function buildOutputModalities(modelType: ModelType, tags: string[]): string[] {
+  const mods: string[] = ["text"];
+  const tagSet = new Set(tags);
+
+  if (modelType === ModelType.Image || tagSet.has("image-generation")) {
+    mods.push("image");
+  } else if (modelType === ModelType.Video) {
+    mods.push("video");
+  }
+
+  return mods;
+}
+
+async function loadExistingModel(filePath: string): Promise<ExistingModel | null> {
+  try {
+    const file = Bun.file(filePath);
+    if (!(await file.exists())) {
+      return null;
+    }
+    const toml = await import(filePath, { with: { type: "toml" } }).then(
+      (mod) => mod.default,
+    );
+    return toml as ExistingModel;
+  } catch (e) {
+    console.warn(`Warning: Failed to parse existing file ${filePath}:`, e);
+    return null;
+  }
+}
+
+function mergeModel(
+  apiModel: z.infer<typeof VercelModel>,
+  existing: ExistingModel | null,
+): MergedModel {
+  const tagSet = new Set(apiModel.tags);
+  const inputModalities = buildInputModalities(apiModel.tags);
+  const outputModalities = buildOutputModalities(apiModel.type, apiModel.tags);
+
+  // Preserve existing values when available (previously manually specified)
+  const name = existing?.name ?? apiModel.name;
+  const attachment = existing?.attachment ?? (tagSet.has("vision") || tagSet.has("file-input"));
+  const reasoning = existing?.reasoning ?? tagSet.has("reasoning");
+  const toolCall = existing?.tool_call ?? tagSet.has("tool-use");
+  const openWeights = existing?.open_weights ?? false;
+  const family = existing?.family ?? inferFamily(apiModel.id, apiModel.name);
+  const structuredOutput = existing?.structured_output;
+  const knowledge = existing?.knowledge;
+  const interleaved = existing?.interleaved;
+  const status = existing?.status;
+
+  // Release date: use API, fallback to existing, then today
+  const releaseDate = apiModel.released
+    ? timestampToDate(apiModel.released)
+    : (existing?.release_date ?? getTodayDate());
+
+  // Preserve existing limits if API returns 0 (indicates missing/invalid data)
+  const contextLimit = apiModel.context_window > 0
+    ? apiModel.context_window
+    : (existing?.limit?.context ?? 0);
+  const outputLimit = apiModel.max_tokens > 0
+    ? apiModel.max_tokens
+    : (existing?.limit?.output ?? 0);
+
+  const merged: MergedModel = {
+    name,
+    family,
+    attachment,
+    reasoning,
+    tool_call: toolCall,
+    temperature: true,
+    release_date: releaseDate,
+    last_updated: getTodayDate(),
+    open_weights: openWeights,
+    ...(structuredOutput !== undefined && { structured_output: structuredOutput }),
+    ...(knowledge && { knowledge }),
+    ...(interleaved !== undefined && { interleaved }),
+    ...(status && { status }),
+    limit: {
+      context: contextLimit,
+      output: outputLimit,
+    },
+    modalities: {
+      input: inputModalities,
+      output: outputModalities,
+    },
+  };
+
+  if (apiModel.pricing) {
+    const inputPrice = apiModel.pricing.input_tiers?.[0]?.cost ?? apiModel.pricing.input;
+    const outputPrice = apiModel.pricing.output_tiers?.[0]?.cost ?? apiModel.pricing.output;
+    const cacheReadPrice = apiModel.pricing.input_cache_read_tiers?.[0]?.cost ?? apiModel.pricing.input_cache_read;
+    const cacheWritePrice = apiModel.pricing.input_cache_write_tiers?.[0]?.cost ?? apiModel.pricing.input_cache_write;
+
+    if (inputPrice && outputPrice) {
+      merged.cost = {
+        input: parseFloat(inputPrice) * 1_000_000,
+        output: parseFloat(outputPrice) * 1_000_000,
+        ...(cacheReadPrice && {
+          cache_read: parseFloat(cacheReadPrice) * 1_000_000,
+        }),
+        ...(cacheWritePrice && {
+          cache_write: parseFloat(cacheWritePrice) * 1_000_000,
+        }),
+      };
+    }
+  }
+
+  return merged;
+}
+
+function formatToml(model: MergedModel): string {
+  const lines: string[] = [];
+
+  lines.push(`name = "${model.name.replace(/"/g, '\\"')}"`);
+  if (model.family) {
+    lines.push(`family = "${model.family}"`);
+  }
+  lines.push(`attachment = ${model.attachment}`);
+  lines.push(`reasoning = ${model.reasoning}`);
+  lines.push(`tool_call = ${model.tool_call}`);
+  if (model.structured_output !== undefined) {
+    lines.push(`structured_output = ${model.structured_output}`);
+  }
+  lines.push(`temperature = ${model.temperature}`);
+  if (model.knowledge) {
+    lines.push(`knowledge = "${model.knowledge}"`);
+  }
+  lines.push(`release_date = "${model.release_date}"`);
+  lines.push(`last_updated = "${model.last_updated}"`);
+  lines.push(`open_weights = ${model.open_weights}`);
+  if (model.status) {
+    lines.push(`status = "${model.status}"`);
+  }
+
+  if (model.interleaved !== undefined) {
+    lines.push("");
+    if (model.interleaved === true) {
+      lines.push(`interleaved = true`);
+    } else if (typeof model.interleaved === "object") {
+      lines.push(`[interleaved]`);
+      lines.push(`field = "${model.interleaved.field}"`);
+    }
+  }
+
+  if (model.cost) {
+    lines.push("");
+    lines.push(`[cost]`);
+    lines.push(`input = ${model.cost.input}`);
+    lines.push(`output = ${model.cost.output}`);
+    if (model.cost.cache_read !== undefined) {
+      lines.push(`cache_read = ${model.cost.cache_read}`);
+    }
+    if (model.cost.cache_write !== undefined) {
+      lines.push(`cache_write = ${model.cost.cache_write}`);
+    }
+  }
+
+  lines.push("");
+  lines.push(`[limit]`);
+  lines.push(`context = ${formatNumber(model.limit.context)}`);
+  lines.push(`output = ${formatNumber(model.limit.output)}`);
+
+  lines.push("");
+  lines.push(`[modalities]`);
+  lines.push(`input = [${model.modalities.input.map((m) => `"${m}"`).join(", ")}]`);
+  lines.push(`output = [${model.modalities.output.map((m) => `"${m}"`).join(", ")}]`);
+
+  return lines.join("\n") + "\n";
+}
+
+function detectChanges(
+  existing: ExistingModel | null,
+  merged: MergedModel,
+): Changes[] {
+  if (!existing) return [];
+
+  const changes: Changes[] = [];
+  const EPSILON = 0.001; // price diff to ignore (per million tokens)
+
+  const shouldSkipZero = (field: string, oldVal: unknown, newVal: unknown): boolean => {
+    if (!Object.values(SkipZeroFields).includes(field as SkipZeroFields)) {
+      return false;
+    }
+    return (typeof oldVal === "number" && oldVal === 0) || (typeof newVal === "number" && newVal === 0);
+  };
+
+  const formatValue = (val: unknown): string => {
+    if (typeof val === "number") return formatNumber(val);
+    if (Array.isArray(val)) return `[${val.join(", ")}]`;
+    if (val === undefined) return "(none)";
+    return String(val);
+  };
+
+  const isMaterialPriceDiff = (oldPrice: unknown, newPrice: unknown): boolean => {
+    // 0 → undefined is not material (cost removed)
+    if (oldPrice === 0 && newPrice === undefined) return false;
+
+    if (oldPrice !== undefined && newPrice !== undefined) {
+      return Math.abs((oldPrice as number) - (newPrice as number)) > EPSILON;
+    }
+
+    return oldPrice !== newPrice;
+  };
+
+  const compare = (field: string, oldVal: unknown, newVal: unknown) => {
+    if (shouldSkipZero(field, oldVal, newVal)) return;
+
+    const isDiff = field.startsWith("cost.")
+      ? isMaterialPriceDiff(oldVal, newVal)
+      : JSON.stringify(oldVal) !== JSON.stringify(newVal);
+
+    if (isDiff) {
+      changes.push({
+        field,
+        oldValue: formatValue(oldVal),
+        newValue: formatValue(newVal),
+      });
+    }
+  };
+
+  compare("name", existing.name, merged.name);
+  compare("family", existing.family, merged.family);
+  compare("attachment", existing.attachment, merged.attachment);
+  compare("reasoning", existing.reasoning, merged.reasoning);
+  compare("tool_call", existing.tool_call, merged.tool_call);
+  compare("structured_output", existing.structured_output, merged.structured_output);
+  compare("open_weights", existing.open_weights, merged.open_weights);
+  compare("release_date", existing.release_date, merged.release_date);
+  compare("cost.input", existing.cost?.input, merged.cost?.input);
+  compare("cost.output", existing.cost?.output, merged.cost?.output);
+  compare("cost.cache_read", existing.cost?.cache_read, merged.cost?.cache_read);
+  compare("cost.cache_write", existing.cost?.cache_write, merged.cost?.cache_write);
+  compare("limit.context", existing.limit?.context, merged.limit.context);
+  compare("limit.output", existing.limit?.output, merged.limit.output);
+  compare("modalities.input", existing.modalities?.input, merged.modalities.input);
+
+  return changes;
+}
+
+async function main() {
+  const args = process.argv.slice(2);
+  const dryRun = args.includes("--dry-run");
+  const newOnly = args.includes("--new-only");
+
+  const modelsDir = path.join(
+    import.meta.dirname,
+    "..",
+    "..",
+    "..",
+    "providers",
+    "vercel",
+    "models",
+  );
+
+  console.log(`${dryRun ? "[DRY RUN] " : ""}${newOnly ? "[NEW ONLY] " : ""}Fetching Vercel models from API...`);
+
+  const res = await fetch(API_ENDPOINT);
+  if (!res.ok) {
+    console.error(`Failed to fetch API: ${res.status} ${res.statusText}`);
+    process.exit(1);
+  }
+
+  const json = await res.json();
+  const parsed = VercelResponse.safeParse(json);
+  if (!parsed.success) {
+    console.error("Invalid API response:", parsed.error.errors);
+    process.exit(1);
+  }
+
+  const apiModels = parsed.data.data;
+
+  const existingFiles = new Set<string>();
+  try {
+    for await (const file of new Bun.Glob("**/*.toml").scan({
+      cwd: modelsDir,
+      absolute: false,
+    })) {
+      existingFiles.add(file);
+    }
+  } catch {
+  }
+
+  console.log(`Found ${apiModels.length} models in API, ${existingFiles.size} existing files\n`);
+
+  const apiModelIds = new Set<string>();
+
+  let created = 0;
+  let updated = 0;
+  let unchanged = 0;
+
+  for (const apiModel of apiModels) {
+    // Skip these since OpenCode does not support image / video generation yet
+    if (apiModel.type === ModelType.Image || apiModel.type === ModelType.Video) {
+      continue;
+    }
+
+    const relativePath = `${apiModel.id}.toml`;
+    const filePath = path.join(modelsDir, relativePath);
+    const dirPath = path.dirname(filePath);
+
+    apiModelIds.add(relativePath);
+
+    const existing = await loadExistingModel(filePath);
+    const merged = mergeModel(apiModel, existing);
+    const tomlContent = formatToml(merged);
+
+    if (existing === null) {
+      created++;
+      if (dryRun) {
+        console.log(`[DRY RUN] Would create: ${relativePath}`);
+        console.log(`  name = "${merged.name}"`);
+        if (merged.family) {
+          console.log(`  family = "${merged.family}" (inferred)`);
+        }
+        console.log("");
+      } else {
+        await mkdir(dirPath, { recursive: true });
+        await Bun.write(filePath, tomlContent);
+        console.log(`Created: ${relativePath}`);
+      }
+    } else {
+      if (newOnly) {
+        unchanged++;
+        continue;
+      }
+
+      const changes = detectChanges(existing, merged);
+
+      if (changes.length > 0) {
+        updated++;
+        if (dryRun) {
+          console.log(`[DRY RUN] Would update: ${relativePath}`);
+        } else {
+          await mkdir(dirPath, { recursive: true });
+          await Bun.write(filePath, tomlContent);
+          console.log(`Updated: ${relativePath}`);
+        }
+        for (const change of changes) {
+          console.log(`  ${change.field}: ${change.oldValue} → ${change.newValue}`);
+        }
+        console.log("");
+      } else {
+        unchanged++;
+      }
+    }
+  }
+
+  const orphaned: string[] = [];
+  for (const file of existingFiles) {
+    if (!apiModelIds.has(file)) {
+      orphaned.push(file);
+      console.log(`Warning: Orphaned file (not in API): ${file}`);
+    }
+  }
+
+  console.log("");
+  if (dryRun) {
+    console.log(
+      `Summary: ${created} would be created, ${updated} would be updated, ${unchanged} unchanged, ${orphaned.length} orphaned`,
+    );
+  } else {
+    console.log(
+      `Summary: ${created} created, ${updated} updated, ${unchanged} unchanged, ${orphaned.length} orphaned`,
+    );
+  }
+}
+
+await main();

packages/core/src/family.ts

@@ -8,6 +8,7 @@ export const ModelFamilyValues = [
   // OpenAI/GPT style
   "gpt",
   "gpt-codex",
+  "gpt-codex-spark",
   "gpt-codex-mini",
   "gpt-pro",
   "gpt-mini",
@@ -127,9 +128,6 @@ export const ModelFamilyValues = [
   "solar-mini",
   "solar-pro",
 
-  // Exaone
-  "exaone",
-
   // Step (StepFun)
   "step",
 
@@ -187,7 +185,6 @@ export const ModelFamilyValues = [
   // Pony
   "pony",
 
-
   // Mercury
   "mercury",
 
@@ -197,6 +194,9 @@ export const ModelFamilyValues = [
   // Mimo
   "mimo",
 
+  // Clarifai
+  "mm-poly",
+
   // Longcat
   "longcat",
 
@@ -284,6 +284,9 @@ export const ModelFamilyValues = [
   // Parakeet
   "parakeet",
 
+  // MiMo
+  "mimo-flash-free",
+
   // NeMo
   "nemoretriever",
 
@@ -364,7 +367,10 @@ export const ModelFamilyValues = [
   "sourceful",
 
   // AllenAI
-  "allenai"
+  "allenai",
+
+  // Writer
+  "palmyra",
 ] as const;
 
 export const ModelFamily = z.enum(ModelFamilyValues);

packages/core/src/schema.ts

@@ -73,6 +73,7 @@ export const Model = z
       .object({
         npm: z.string().optional(),
         api: z.string().optional(),
+        shape: z.enum(["responses", "completions"]).optional(),
       })
       .optional(),
   })

providers/abacus/models/claude-opus-4-6.toml

@@ -0,0 +1,22 @@
+name = "Claude Opus 4.6"
+family = "claude-opus"
+release_date = "2026-02-05"
+last_updated = "2026-02-05"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-05"
+open_weights = false
+
+[cost]
+input = 5.00
+output = 25.00
+
+[limit]
+context = 200_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/abacus/models/claude-sonnet-4-6.toml

@@ -0,0 +1,22 @@
+name = "Claude Sonnet 4.6"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/abacus/models/deepseek/deepseek-v3.1.toml

@@ -9,8 +9,8 @@ tool_call = true
 open_weights = true
 
 [cost]
-input = 0.14
-output = 0.28
+input = 0.55
+output = 1.66
 
 [limit]
 context = 128_000

providers/abacus/models/gemini-3.1-flash-lite-preview.toml

@@ -0,0 +1,22 @@
+name = "Gemini 3.1 Flash Lite Preview"
+family = "gemini-flash"
+release_date = "2026-03-01"
+last_updated = "2026-03-01"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.25
+output = 1.50
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]

providers/abacus/models/gemini-3.1-pro-preview.toml

@@ -0,0 +1,23 @@
+name = "Gemini 3.1 Pro Preview"
+family = "gemini-pro"
+release_date = "2026-02-19"
+last_updated = "2026-02-19"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-01"
+open_weights = false
+
+[cost]
+input = 2.00
+output = 12.00
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video", "audio", "pdf"]
+output = ["text"]

providers/abacus/models/gpt-5-codex.toml

@@ -0,0 +1,24 @@
+name = "GPT-5 Codex"
+family = "gpt"
+release_date = "2025-09-15"
+last_updated = "2025-09-15"
+attachment = false
+reasoning = true
+temperature = false
+knowledge = "2024-09-30"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 1.25
+output = 10.00
+
+[limit]
+context = 400_000
+input = 272_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/abacus/models/gpt-5.1-codex-max.toml

@@ -1,7 +1,7 @@
-name = "GPT-5"
+name = "GPT-5.1 Codex Max"
 family = "gpt"
-release_date = "2025-08-07"
-last_updated = "2025-08-07"
+release_date = "2025-11-13"
+last_updated = "2025-11-13"
 attachment = true
 reasoning = true
 temperature = false
@@ -12,11 +12,11 @@ open_weights = false
 
 [cost]
 input = 1.25
-output = 10
-cache_read = 0.13
+output = 10.00
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/abacus/models/gpt-5.1-codex.toml

@@ -0,0 +1,24 @@
+name = "GPT-5.1 Codex"
+family = "gpt"
+release_date = "2025-11-13"
+last_updated = "2025-11-13"
+attachment = true
+reasoning = true
+temperature = false
+knowledge = "2024-09-30"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 1.25
+output = 10.00
+
+[limit]
+context = 400_000
+input = 272_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/abacus/models/gpt-5.2-chat-latest.toml

@@ -10,8 +10,8 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 1.50
-output = 12.00
+input = 1.75
+output = 14.00
 
 [limit]
 context = 400_000

providers/abacus/models/gpt-5.2-codex.toml

@@ -0,0 +1,24 @@
+name = "GPT-5.2 Codex"
+family = "gpt"
+release_date = "2025-12-11"
+last_updated = "2025-12-11"
+attachment = true
+reasoning = true
+temperature = false
+knowledge = "2025-08-31"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14.00
+
+[limit]
+context = 400_000
+input = 272_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/abacus/models/gpt-5.3-chat-latest.toml

@@ -0,0 +1,21 @@
+name = "GPT-5.3 Chat Latest"
+family = "gpt"
+release_date = "2026-03-01"
+last_updated = "2026-03-01"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14.00
+
+[limit]
+context = 400_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/abacus/models/gpt-5.3-codex-xhigh.toml

@@ -0,0 +1,24 @@
+name = "GPT-5.3 Codex XHigh"
+family = "gpt"
+release_date = "2026-02-05"
+last_updated = "2026-02-05"
+attachment = true
+reasoning = true
+temperature = false
+knowledge = "2025-08-31"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14.00
+
+[limit]
+context = 400_000
+input = 272_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/abacus/models/gpt-5.3-codex.toml

@@ -0,0 +1,24 @@
+name = "GPT-5.3 Codex"
+family = "gpt"
+release_date = "2026-02-05"
+last_updated = "2026-02-05"
+attachment = true
+reasoning = true
+temperature = false
+knowledge = "2025-08-31"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14.00
+
+[limit]
+context = 400_000
+input = 272_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/abacus/models/kimi-k2.5.toml

@@ -0,0 +1,23 @@
+name = "Kimi K2.5"
+family = "kimi"
+release_date = "2026-01"
+last_updated = "2026-01"
+attachment = false
+reasoning = true
+structured_output = true
+temperature = true
+tool_call = true
+knowledge = "2025-01"
+open_weights = true
+
+[cost]
+input = 0.60
+output = 3.00
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/abacus/models/o3-pro.toml

@@ -11,7 +11,7 @@ open_weights = false
 
 [cost]
 input = 20.00
-output = 80.00
+output = 40.00
 
 [limit]
 context = 200_000

providers/abacus/models/route-llm.toml

@@ -10,8 +10,8 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0.50
-output = 1.50
+input = 3.00
+output = 15.00
 
 [limit]
 context = 128_000

providers/abacus/models/zai-org/glm-4.7.toml

@@ -9,8 +9,8 @@ tool_call = true
 open_weights = true
 
 [cost]
-input = 0.70
-output = 2.50
+input = 0.60
+output = 2.20
 
 [limit]
 context = 128_000

providers/abacus/models/zai-org/glm-5.toml

@@ -0,0 +1,21 @@
+name = "GLM-5"
+family = "glm"
+release_date = "2026-02-11"
+last_updated = "2026-02-11"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[cost]
+input = 1.00
+output = 3.20
+
+[limit]
+context = 204800
+output = 131072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/claude-opus-4-6-think.toml

@@ -0,0 +1,30 @@
+name = "Claude Opus 4.6 Think"
+family = "claude-opus"
+release_date = "2026-02-05"
+last_updated = "2026-02-05"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-05"
+open_weights = false
+
+[cost]
+input = 5.00
+output = 25.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.00
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/aihubmix/models/claude-opus-4-6.toml

@@ -0,0 +1,30 @@
+name = "Claude Opus 4.6"
+family = "claude-opus"
+release_date = "2026-02-05"
+last_updated = "2026-02-05"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-05"
+open_weights = false
+
+[cost]
+input = 5.00
+output = 25.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.00
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/aihubmix/models/claude-sonnet-4-6-think.toml

@@ -0,0 +1,30 @@
+name = "Claude Sonnet 4.6 Think"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.50
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/aihubmix/models/claude-sonnet-4-6.toml

@@ -0,0 +1,30 @@
+name = "Claude Sonnet 4.6"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.50
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/aihubmix/models/coding-glm-5-free.toml

@@ -0,0 +1,24 @@
+name = "Coding-GLM-5-Free"
+family = "glm"
+release_date = "2026-02-11"
+last_updated = "2026-02-11"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.0
+output = 0.0
+
+[limit]
+context = 204800
+output = 131072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/gemini-3-pro-preview-search.toml

@@ -0,0 +1,23 @@
+name = "Gemini 3 Pro Preview Search"
+family = "gemini-pro"
+release_date = "2025-11-19"
+last_updated = "2025-11-19"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-11"
+open_weights = false
+
+[cost]
+input = 2.00
+output = 12.00
+cache_read = 0.50
+
+[limit]
+context = 1_000_000
+output = 65_000
+
+[modalities]
+input = ["text", "image", "audio", "video"]
+output = ["text"]

providers/aihubmix/models/glm-5.toml

@@ -0,0 +1,24 @@
+name = "GLM-5"
+family = "glm"
+release_date = "2026-02-11"
+last_updated = "2026-02-11"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.88
+output = 2.82
+
+[limit]
+context = 204800
+output = 131072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/gpt-5.2-codex.toml

@@ -0,0 +1,24 @@
+name = "GPT-5.2-Codex"
+family = "gpt-codex"
+release_date = "2026-01-14"
+last_updated = "2026-01-14"
+attachment = true
+reasoning = true
+temperature = true
+knowledge = "2025-08-31"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14.00
+cache_read = 0.175
+
+[limit]
+context = 400_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/aihubmix/models/minimax-m2.5.toml

@@ -0,0 +1,21 @@
+name = "MiniMax-M2.5"
+family = "minimax"
+release_date = "2026-02-12"
+last_updated = "2026-02-12"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.29
+output = 1.15
+
+[limit]
+context = 204_800
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/qwen3-coder-next.toml

@@ -0,0 +1,23 @@
+name = "Qwen3 Coder Next"
+family = "qwen"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = false
+temperature = true
+release_date = "2026-02-04"
+last_updated = "2026-02-04"
+open_weights = true
+
+[cost]
+input = 0.14
+output = 0.55
+
+[limit]
+context = 262_144
+input = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/qwen3.5-plus.toml

@@ -0,0 +1,22 @@
+name = "Qwen 3.5 Plus"
+family = "qwen"
+release_date = "2026-02-16"
+last_updated = "2026-02-16"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0.11
+output = 0.66
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba-cn/models/MiniMax/MiniMax-M2.5.toml

@@ -0,0 +1,21 @@
+name = "MiniMax M2.5"
+family = "minimax"
+release_date = "2026-02-12"
+last_updated = "2026-02-12"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.301
+output = 1.205
+
+[limit]
+context = 204_800
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-cn/models/glm-5.toml

@@ -0,0 +1,24 @@
+name = "GLM-5"
+family = "glm"
+release_date = "2026-02-11"
+last_updated = "2026-02-11"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.86
+output = 3.15
+
+[limit]
+context = 202_752
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-cn/models/kimi-k2.5.toml

@@ -1,12 +1,13 @@
 name = "Moonshot Kimi K2.5"
 family = "kimi"
-release_date = "2025-01-27"
-last_updated = "2025-01-27"
+release_date = "2026-01-27"
+last_updated = "2026-01-27"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
 open_weights = true
+knowledge = "2025-01"
 structured_output = false
 
 [interleaved]
@@ -21,5 +22,5 @@ context = 262_144
 output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video"]
 output = ["text"]

providers/alibaba-cn/models/kimi/kimi-k2.5.toml

@@ -0,0 +1,27 @@
+name = "kimi/kimi-k2.5"
+family = "kimi"
+release_date = "2026-01-27"
+last_updated = "2026-01-27"
+attachment = false
+reasoning = true
+structured_output = true
+temperature = false
+tool_call = true
+knowledge = "2025-01"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.6
+output = 3.0
+cache_read = 0.1
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba-cn/models/minimax-m2.5.toml

@@ -0,0 +1,24 @@
+name = "MiniMax-M2.5"
+family = "minimax"
+release_date = "2026-02-12"
+last_updated = "2026-02-12"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.30
+output = 1.20
+
+[interleaved]
+field = "reasoning_content"
+
+[limit]
+context = 204_800
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-cn/models/qwen3.5-397b-a17b.toml

@@ -0,0 +1,23 @@
+name = "Qwen3.5 397B-A17B"
+family = "qwen"
+release_date = "2026-02-16"
+last_updated = "2026-02-16"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.430
+output = 2.58
+reasoning = 2.58
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba-cn/models/qwen3.5-flash.toml

@@ -0,0 +1,24 @@
+name = "Qwen3.5 Flash"
+family = "qwen"
+release_date = "2026-02-23"
+last_updated = "2026-02-23"
+attachment = true
+reasoning = true
+structured_output = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0.172
+output = 1.72
+reasoning = 1.72
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba-cn/models/qwen3.5-plus.toml

@@ -0,0 +1,23 @@
+name = "Qwen3.5 Plus"
+family = "qwen"
+release_date = "2026-02-16"
+last_updated = "2026-02-16"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0.573
+output = 3.44
+reasoning = 3.44
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba-cn/models/siliconflow/deepseek-r1-0528.toml

@@ -1,6 +1,6 @@
-name = "MiniMaxAI/MiniMax-M1-80k"
-family = "minimax"
-release_date = "2025-06-17"
+name = "siliconflow/deepseek-r1-0528"
+family = "deepseek-thinking"
+release_date = "2025-05-28"
 last_updated = "2025-11-25"
 attachment = false
 reasoning = true
@@ -10,13 +10,13 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.55
-output = 2.2
+input = 0.5
+output = 2.18
 
 [limit]
-context = 131_000
-output = 131_000
+context = 163_840
+output = 32_768
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/alibaba-cn/models/siliconflow/deepseek-v3-0324.toml

@@ -1,6 +1,6 @@
-name = "moonshotai/Kimi-Dev-72B"
-family = "kimi"
-release_date = "2025-06-19"
+name = "siliconflow/deepseek-v3-0324"
+family = "deepseek"
+release_date = "2024-12-26"
 last_updated = "2025-11-25"
 attachment = false
 reasoning = false
@@ -10,13 +10,13 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.29
-output = 1.15
+input = 0.25
+output = 1.0
 
 [limit]
-context = 131_000
-output = 131_000
+context = 163_840
+output = 163_840
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/alibaba-cn/models/siliconflow/deepseek-v3.1-terminus.toml

@@ -1,6 +1,6 @@
-name = "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B"
-family = "qwen"
-release_date = "2025-01-20"
+name = "siliconflow/deepseek-v3.1-terminus"
+family = "deepseek"
+release_date = "2025-09-29"
 last_updated = "2025-11-25"
 attachment = false
 reasoning = true
@@ -10,13 +10,13 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0.05
-output = 0.05
+input = 0.27
+output = 1.0
 
 [limit]
-context = 33_000
-output = 16_000
+context = 163_840
+output = 65_536
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/alibaba-cn/models/siliconflow/deepseek-v3.2.toml

@@ -0,0 +1,22 @@
+name = "siliconflow/deepseek-v3.2"
+family = "deepseek"
+release_date = "2025-12-03"
+last_updated = "2025-12-03"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.27
+output = 0.42
+
+[limit]
+context = 163_840
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan-cn/logo.svg

@@ -0,0 +1,3 @@
+<svg width="24" height="24" viewBox="0 0 40 40" xmlns="http://www.w3.org/2000/svg">
+<path d="M37.9998 23.021C33.7998 25.2889 29.5698 27.3649 24.8614 28.3069C23.8114 28.5154 22.6474 28.5154 21.5809 28.3714C20.5639 28.2439 20.0554 27.3484 20.4169 26.4064C20.7619 25.5289 21.2209 24.635 21.8119 23.9C23.0899 22.3025 24.5329 20.849 25.8289 19.268C26.6203 18.2991 27.3335 17.2689 27.9618 16.187C28.4208 15.4205 28.2078 14.4935 27.4038 14.111C26.0584 13.4556 24.6154 12.9936 23.1889 12.4986C23.0239 12.4341 22.7779 12.6096 22.4509 12.7221C22.8604 13.0881 23.1559 13.3596 23.5654 13.727C19.3339 14.447 15.3305 15.467 11.4455 16.874C11.4275 16.9535 11.396 17.0165 11.411 17.0495C11.9855 17.927 11.723 18.5975 10.886 19.1405C10.5611 19.3531 10.2732 19.6176 10.034 19.9235C12.593 20.6735 14.873 20.243 17.0539 18.821C16.9234 18.6305 16.7914 18.455 16.6609 18.263C17.4799 18.407 17.9719 18.854 18.0379 19.556C18.0544 19.7165 17.9569 19.8755 17.9074 20.036C17.7919 19.907 17.6449 19.781 17.5474 19.6355C17.4799 19.5395 17.4634 19.4285 17.4154 19.268C14.8235 20.993 12.035 21.425 8.96751 20.531C8.96751 21.137 8.93451 21.6485 8.98401 22.1435C9.01701 22.574 8.83701 22.766 8.44401 22.9895C7.55752 23.5325 6.63803 24.092 5.90003 24.8105C5.01504 25.6879 5.34354 26.7589 6.54053 27.2059C7.90102 27.7159 9.329 27.7309 10.7555 27.5569C12.4445 27.3484 14.1005 27.0769 15.9394 26.8219C13.79 27.8269 11.6735 28.5319 9.4445 28.8169C7.88452 29.0269 6.32753 29.1379 4.78554 28.6909C2.57156 28.0684 1.58607 26.4394 2.16057 24.251C2.70206 22.2065 4.01455 20.5775 5.42454 19.076C10.133 14.078 16.0864 11.5401 22.9744 11.0286C24.5824 10.9176 26.2069 11.1246 27.7143 11.7951C29.8308 12.7536 30.7173 14.78 29.6838 16.826C29.0118 18.1835 28.0758 19.4285 27.1413 20.6585C26.2234 21.872 25.1899 22.9895 24.2224 24.155C23.9434 24.506 23.6809 24.875 23.4679 25.2724C23.0569 26.0224 23.3359 26.5174 24.2059 26.4394C26.0254 26.2624 27.8808 26.1199 29.6358 25.6729C32.2098 25.0174 34.7193 24.092 37.2618 23.2775C37.5243 23.213 37.7703 23.117 37.9998 23.0225V23.021Z" fill="currentColor"/>
+</svg>

providers/alibaba-coding-plan-cn/models/glm-4.7.toml

@@ -0,0 +1,27 @@
+name = "GLM-4.7"
+family = "glm"
+release_date = "2025-12-22"
+last_updated = "2025-12-22"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 202_752
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan-cn/models/glm-5.toml

@@ -0,0 +1,26 @@
+name = "GLM-5"
+family = "glm"
+release_date = "2026-02-11"
+last_updated = "2026-02-11"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 202_752
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan-cn/models/kimi-k2.5.toml

@@ -0,0 +1,27 @@
+name = "Kimi K2.5"
+family = "kimi"
+release_date = "2026-01-27"
+last_updated = "2026-01-27"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-01"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/alibaba-coding-plan-cn/models/qwen3-coder-next.toml

@@ -0,0 +1,24 @@
+name = "Qwen3 Coder Next"
+family = "qwen"
+release_date = "2026-02-03"
+last_updated = "2026-02-03"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan-cn/models/qwen3-coder-plus.toml

@@ -0,0 +1,24 @@
+name = "Qwen3 Coder Plus"
+family = "qwen"
+release_date = "2025-07-23"
+last_updated = "2025-07-23"
+attachment = false
+reasoning = false
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan-cn/models/qwen3-max-2026-01-23.toml

@@ -0,0 +1,24 @@
+name = "Qwen3 Max"
+family = "qwen"
+release_date = "2026-01-23"
+last_updated = "2026-01-23"
+attachment = false
+reasoning = false
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan-cn/models/qwen3.5-plus.toml

@@ -0,0 +1,24 @@
+name = "Qwen3.5 Plus"
+family = "qwen"
+release_date = "2026-02-16"
+last_updated = "2026-02-16"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/alibaba-coding-plan-cn/provider.toml

@@ -0,0 +1,5 @@
+name = "Alibaba Coding Plan (China)"
+env = ["ALIBABA_CODING_PLAN_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+doc = "https://help.aliyun.com/zh/model-studio/coding-plan"
+api = "https://coding.dashscope.aliyuncs.com/v1"

providers/alibaba-coding-plan/logo.svg

@@ -0,0 +1,3 @@
+<svg width="24" height="24" viewBox="0 0 40 40" xmlns="http://www.w3.org/2000/svg">
+<path d="M37.9998 23.021C33.7998 25.2889 29.5698 27.3649 24.8614 28.3069C23.8114 28.5154 22.6474 28.5154 21.5809 28.3714C20.5639 28.2439 20.0554 27.3484 20.4169 26.4064C20.7619 25.5289 21.2209 24.635 21.8119 23.9C23.0899 22.3025 24.5329 20.849 25.8289 19.268C26.6203 18.2991 27.3335 17.2689 27.9618 16.187C28.4208 15.4205 28.2078 14.4935 27.4038 14.111C26.0584 13.4556 24.6154 12.9936 23.1889 12.4986C23.0239 12.4341 22.7779 12.6096 22.4509 12.7221C22.8604 13.0881 23.1559 13.3596 23.5654 13.727C19.3339 14.447 15.3305 15.467 11.4455 16.874C11.4275 16.9535 11.396 17.0165 11.411 17.0495C11.9855 17.927 11.723 18.5975 10.886 19.1405C10.5611 19.3531 10.2732 19.6176 10.034 19.9235C12.593 20.6735 14.873 20.243 17.0539 18.821C16.9234 18.6305 16.7914 18.455 16.6609 18.263C17.4799 18.407 17.9719 18.854 18.0379 19.556C18.0544 19.7165 17.9569 19.8755 17.9074 20.036C17.7919 19.907 17.6449 19.781 17.5474 19.6355C17.4799 19.5395 17.4634 19.4285 17.4154 19.268C14.8235 20.993 12.035 21.425 8.96751 20.531C8.96751 21.137 8.93451 21.6485 8.98401 22.1435C9.01701 22.574 8.83701 22.766 8.44401 22.9895C7.55752 23.5325 6.63803 24.092 5.90003 24.8105C5.01504 25.6879 5.34354 26.7589 6.54053 27.2059C7.90102 27.7159 9.329 27.7309 10.7555 27.5569C12.4445 27.3484 14.1005 27.0769 15.9394 26.8219C13.79 27.8269 11.6735 28.5319 9.4445 28.8169C7.88452 29.0269 6.32753 29.1379 4.78554 28.6909C2.57156 28.0684 1.58607 26.4394 2.16057 24.251C2.70206 22.2065 4.01455 20.5775 5.42454 19.076C10.133 14.078 16.0864 11.5401 22.9744 11.0286C24.5824 10.9176 26.2069 11.1246 27.7143 11.7951C29.8308 12.7536 30.7173 14.78 29.6838 16.826C29.0118 18.1835 28.0758 19.4285 27.1413 20.6585C26.2234 21.872 25.1899 22.9895 24.2224 24.155C23.9434 24.506 23.6809 24.875 23.4679 25.2724C23.0569 26.0224 23.3359 26.5174 24.2059 26.4394C26.0254 26.2624 27.8808 26.1199 29.6358 25.6729C32.2098 25.0174 34.7193 24.092 37.2618 23.2775C37.5243 23.213 37.7703 23.117 37.9998 23.0225V23.021Z" fill="currentColor"/>
+</svg>

providers/alibaba-coding-plan/models/MiniMax-M2.5.toml

@@ -0,0 +1,26 @@
+name = "MiniMax-M2.5"
+family = "minimax"
+release_date = "2026-02-12"
+last_updated = "2026-02-12"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 204_800
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan/models/glm-4.7.toml

@@ -0,0 +1,27 @@
+name = "GLM-4.7"
+family = "glm"
+release_date = "2025-12-22"
+last_updated = "2025-12-22"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 202_752
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan/models/glm-5.toml

@@ -0,0 +1,26 @@
+name = "GLM-5"
+family = "glm"
+release_date = "2026-02-11"
+last_updated = "2026-02-11"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 202_752
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan/models/kimi-k2.5.toml

@@ -0,0 +1,27 @@
+name = "Kimi K2.5"
+family = "kimi"
+release_date = "2026-01-27"
+last_updated = "2026-01-27"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-01"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 262_144
+output = 32_768
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba-coding-plan/models/qwen3-coder-next.toml

@@ -0,0 +1,24 @@
+name = "Qwen3 Coder Next"
+family = "qwen"
+release_date = "2026-02-03"
+last_updated = "2026-02-03"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan/models/qwen3-coder-plus.toml

@@ -0,0 +1,24 @@
+name = "Qwen3 Coder Plus"
+family = "qwen"
+release_date = "2025-07-23"
+last_updated = "2025-07-23"
+attachment = false
+reasoning = false
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan/models/qwen3-max-2026-01-23.toml

@@ -0,0 +1,24 @@
+name = "Qwen3 Max"
+family = "qwen"
+release_date = "2026-01-23"
+last_updated = "2026-01-23"
+attachment = false
+reasoning = false
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/alibaba-coding-plan/models/qwen3.5-plus.toml

@@ -0,0 +1,24 @@
+name = "Qwen3.5 Plus"
+family = "qwen"
+release_date = "2026-02-16"
+last_updated = "2026-02-16"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba-coding-plan/provider.toml

@@ -0,0 +1,5 @@
+name = "Alibaba Coding Plan"
+env = ["ALIBABA_CODING_PLAN_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+doc = "https://www.alibabacloud.com/help/en/model-studio/coding-plan"
+api = "https://coding-intl.dashscope.aliyuncs.com/v1"

providers/alibaba/models/qwen3.5-397b-a17b.toml

@@ -0,0 +1,23 @@
+name = "Qwen3.5 397B-A17B"
+family = "qwen"
+release_date = "2026-02-16"
+last_updated = "2026-02-16"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.6
+output = 3.6
+reasoning = 3.6
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/alibaba/models/qwen3.5-plus.toml

@@ -0,0 +1,24 @@
+name = "Qwen3.5 Plus"
+family = "qwen"
+release_date = "2026-02-16"
+last_updated = "2026-02-16"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2025-04"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0.4
+output = 2.4
+reasoning = 2.4
+
+
+[limit]
+context = 1_000_000
+output = 65_536
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/amazon-bedrock/models/anthropic.claude-opus-4-6-v1.toml

@@ -22,7 +22,7 @@ cache_read = 1.00
 cache_write = 12.50
 
 [limit]
-context = 1_000_000
+context = 200_000
 output = 128_000
 
 [modalities]

providers/amazon-bedrock/models/anthropic.claude-sonnet-4-6.toml

@@ -0,0 +1,30 @@
+name = "Claude Sonnet 4.6"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.50
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/amazon-bedrock/models/deepseek.v3.2-v1:0.toml

@@ -0,0 +1,22 @@
+name = "DeepSeek-V3.2"
+family = "deepseek"
+release_date = "2026-02-15"
+last_updated = "2026-02-15"
+attachment = false
+reasoning = true
+temperature = true
+knowledge = "2024-07"
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.62
+output = 1.85
+
+[limit]
+context = 163_840
+output = 81_920
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/amazon-bedrock/models/eu.anthropic.claude-opus-4-6-v1.toml

@@ -22,7 +22,7 @@ cache_read = 1.00
 cache_write = 12.50
 
 [limit]
-context = 1_000_000
+context = 200_000
 output = 128_000
 
 [modalities]

providers/amazon-bedrock/models/eu.anthropic.claude-sonnet-4-6.toml

@@ -0,0 +1,30 @@
+name = "Claude Sonnet 4.6 (EU)"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.50
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/amazon-bedrock/models/global.anthropic.claude-opus-4-6-v1.toml

@@ -22,7 +22,7 @@ cache_read = 1.00
 cache_write = 12.50
 
 [limit]
-context = 1_000_000
+context = 200_000
 output = 128_000
 
 [modalities]

providers/amazon-bedrock/models/global.anthropic.claude-sonnet-4-6.toml

@@ -0,0 +1,30 @@
+name = "Claude Sonnet 4.6 (Global)"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.50
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/amazon-bedrock/models/minimax.minimax-m2.1.toml

@@ -0,0 +1,22 @@
+name = "MiniMax M2.1"
+family = "minimax"
+release_date = "2025-12-23"
+last_updated = "2025-12-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.30
+output = 1.20
+
+[limit]
+context = 204_800
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/amazon-bedrock/models/mistral.devstral-2-123b

@@ -0,0 +1,21 @@
+name = "Devstral 2 135B"
+family = "mistral"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0.40
+output = 2.00
+
+[limit]
+context = 256_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/amazon-bedrock/models/us.anthropic.claude-opus-4-6-v1.toml

@@ -22,7 +22,7 @@ cache_read = 1.00
 cache_write = 12.50
 
 [limit]
-context = 1_000_000
+context = 200_000
 output = 128_000
 
 [modalities]

providers/amazon-bedrock/models/us.anthropic.claude-sonnet-4-6.toml

@@ -0,0 +1,30 @@
+name = "Claude Sonnet 4.6 (US)"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.50
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/amazon-bedrock/models/writer.palmyra-x4-v1:0.toml

@@ -0,0 +1,21 @@
+name = "Palmyra X4"
+family = "palmyra"
+release_date = "2025-04-28"
+last_updated = "2025-04-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[cost]
+input = 2.5
+output = 10
+
+[limit]
+context = 122_880
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/amazon-bedrock/models/writer.palmyra-x5-v1:0.toml

@@ -0,0 +1,21 @@
+name = "Palmyra X5"
+family = "palmyra"
+release_date = "2025-04-28"
+last_updated = "2025-04-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[cost]
+input = 0.6
+output = 6
+
+[limit]
+context = 1_040_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/amazon-bedrock/models/zai.glm-4.7-flash.toml

@@ -0,0 +1,22 @@
+name = "GLM-4.7-Flash"
+family = "glm-flash"
+release_date = "2026-01-19"
+last_updated = "2026-01-19"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+
+[cost]
+input = 0.07
+output = 0.40
+
+[limit]
+context = 200_000
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/amazon-bedrock/models/zai.glm-4.7.toml

@@ -1,7 +1,7 @@
-name = "GLM 4.6"
+name = "GLM-4.7"
 family = "glm"
-release_date = "2025-10-01"
-last_updated = "2025-10-01"
+release_date = "2025-12-22"
+last_updated = "2025-12-22"
 attachment = false
 reasoning = true
 temperature = true
@@ -9,18 +9,17 @@ tool_call = true
 knowledge = "2025-04"
 open_weights = true
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
-input = 0.55
-output = 2.19
-cache_read = 0.28
+input = 0.60
+output = 2.20
 
 [limit]
-context = 198_000
-output = 198_000
+context = 204_800
+output = 131_072
 
 [modalities]
 input = ["text"]
 output = ["text"]
-
-[interleaved]
-field = "reasoning_content"

providers/amazon-bedrock/provider.toml

@@ -1,4 +1,4 @@
 name = "Amazon Bedrock"
-env = ["AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY", "AWS_REGION"]
+env = ["AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY", "AWS_REGION", "AWS_BEARER_TOKEN_BEDROCK"]
 npm = "@ai-sdk/amazon-bedrock"
 doc = "https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html"

providers/anthropic/models/claude-sonnet-4-6.toml

@@ -0,0 +1,30 @@
+name = "Claude Sonnet 4.6"
+family = "claude-sonnet"
+release_date = "2026-02-17"
+last_updated = "2026-02-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-08"
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[cost.context_over_200k]
+input = 6.00
+output = 22.50
+cache_read = 0.60
+cache_write = 7.50
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/azure-cognitive-services/models/claude-haiku-4-5.toml

@@ -1 +0,0 @@
-../../azure/models/claude-haiku-4-5.toml

providers/azure-cognitive-services/models/claude-haiku-4-5.toml

@@ -0,0 +1,29 @@
+name = "Claude Haiku 4.5"
+family = "claude-haiku"
+release_date = "2025-11-18"
+last_updated = "2025-11-18"
+attachment = true
+reasoning = true
+temperature = true
+knowledge = "2025-02-31"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 1.00
+output = 5.00
+cache_read = 0.10
+cache_write = 1.25
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"

providers/azure-cognitive-services/models/claude-opus-4-1.toml

@@ -1 +0,0 @@
-../../azure/models/claude-opus-4-1.toml

providers/azure-cognitive-services/models/claude-opus-4-1.toml

@@ -0,0 +1,29 @@
+name = "Claude Opus 4.1"
+family = "claude-opus"
+release_date = "2025-11-18"
+last_updated = "2025-11-18"
+attachment = true
+reasoning = true
+temperature = true
+knowledge = "2025-03-31"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 15.00
+output = 75.00
+cache_read = 1.50
+cache_write = 18.75
+
+[limit]
+context = 200_000
+output = 32_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"

providers/azure-cognitive-services/models/claude-opus-4-5.toml

@@ -1 +0,0 @@
-../../azure/models/claude-opus-4-5.toml

providers/azure-cognitive-services/models/claude-opus-4-5.toml

@@ -0,0 +1,28 @@
+name = "Claude Opus 4.5"
+family = "claude-opus"
+release_date = "2025-11-24"
+last_updated = "2025-08-01"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-03-31"
+open_weights = false
+
+[cost]
+input = 5.00
+output = 25.00
+cache_read = 0.50
+cache_write = 6.25
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"

providers/azure-cognitive-services/models/claude-opus-4-6.toml

@@ -1 +0,0 @@
-../../azure/models/claude-opus-4-6.toml

providers/azure-cognitive-services/models/claude-opus-4-6.toml

@@ -0,0 +1,34 @@
+name = "Claude Opus 4.6"
+family = "claude-opus"
+release_date = "2026-02-05"
+last_updated = "2026-02-05"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-05"
+open_weights = false
+
+[cost]
+input = 5.00
+output = 25.00
+cache_read = 0.50
+cache_write = 6.25
+
+[cost.context_over_200k]
+input = 10.00
+output = 37.50
+cache_read = 1.00
+cache_write = 12.50
+
+[limit]
+context = 200_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"

providers/azure-cognitive-services/models/claude-sonnet-4-5.toml

@@ -1 +0,0 @@
-../../azure/models/claude-sonnet-4-5.toml

providers/azure-cognitive-services/models/claude-sonnet-4-5.toml

@@ -0,0 +1,29 @@
+name = "Claude Sonnet 4.5"
+family = "claude-sonnet"
+release_date = "2025-11-18"
+last_updated = "2025-11-18"
+attachment = true
+reasoning = true
+temperature = true
+knowledge = "2025-07-31"
+tool_call = true
+structured_output = true
+open_weights = false
+
+[cost]
+input = 3.00
+output = 15.00
+cache_read = 0.30
+cache_write = 3.75
+
+[limit]
+context = 200_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"

providers/azure-cognitive-services/models/gpt-5.2.toml

@@ -0,0 +1 @@
+../../azure/models/gpt-5.2.toml

providers/azure-cognitive-services/models/gpt-5.3-codex.toml

@@ -0,0 +1 @@
+../../azure/models/gpt-5.3-codex.toml