use Venice context tier thresholds

feat(chutes): add API-driven model generator script

bun.lock

@@ -5,7 +5,7 @@
     "": {
       "name": "models.dev",
       "dependencies": {
-        "@cloudflare/workers-types": "^4.20250801.0",
+        "@cloudflare/workers-types": "^4.20260424.1",
         "sst": "3.17.23",
       },
     },
@@ -49,7 +49,7 @@
     "zod": "3.24.2",
   },
   "packages": {
-    "@cloudflare/workers-types": ["@cloudflare/workers-types@4.20250801.0", "", {}, "sha512-BQmMdoOGClY23TesgkR1PeGrPvPsSFD/zW7pDzWZHkOEsqkPk2A91h52bP8GbtKYTl1vdaYjQgJlGsP6Ih4G0w=="],
+    "@cloudflare/workers-types": ["@cloudflare/workers-types@4.20260424.1", "", {}, "sha512-0DLJ9yEk1KKzPbqop80Gw/P1wkKKzawmipULiJWdBXIBCoMvE0OVWms3IrL/Q/G7tfmPop9yF4XlZ69k9JLYng=="],
 
     "@modelcontextprotocol/sdk": ["@modelcontextprotocol/sdk@1.6.1", "", { "dependencies": { "content-type": "^1.0.5", "cors": "^2.8.5", "eventsource": "^3.0.2", "express": "^5.0.1", "express-rate-limit": "^7.5.0", "pkce-challenge": "^4.1.0", "raw-body": "^3.0.0", "zod": "^3.23.8", "zod-to-json-schema": "^3.24.1" } }, "sha512-oxzMzYCkZHMntzuyerehK3fV6A2Kwh5BD6CGEJSVDU2QNEhfLOptf2X7esQgaHZXHZY0oHmMsOtIDLP71UJXgA=="],
 
@@ -59,7 +59,7 @@
 
     "@tsconfig/bun": ["@tsconfig/bun@1.0.8", "", {}, "sha512-JlJaRaS4hBTypxtFe8WhnwV8blf0R+3yehLk8XuyxUYNx6VXsKCjACSCvOYEFUiqlhlBWxtYCn/zRlOb8BzBQg=="],
 
-    "@types/bun": ["@types/bun@1.2.16", "", { "dependencies": { "bun-types": "1.2.16" } }, "sha512-1aCZJ/6nSiViw339RsaNhkNoEloLaPzZhxMOYEa7OzRzO41IGg5n/7I43/ZIAW/c+Q6cT12Vf7fOZOoVIzb5BQ=="],
+    "@types/bun": ["@types/bun@1.3.0", "", { "dependencies": { "bun-types": "1.3.0" } }, "sha512-+lAGCYjXjip2qY375xX/scJeVRmZ5cY0wyHYyCYxNcdEXrQ4AOe3gACgd4iQ8ksOslJtW4VNxBJ8llUwc3a6AA=="],
 
     "@types/node": ["@types/node@22.13.9", "", { "dependencies": { "undici-types": "~6.20.0" } }, "sha512-acBjXdRJ3A6Pb3tqnw9HZmyR3Fiol3aGxRCK1x3d+6CDAMjl7I649wpSd+yNURCjbOUGu9tqtLKnTGxmK6CyGw=="],
 
@@ -79,7 +79,7 @@
 
     "buffer": ["buffer@4.9.2", "", { "dependencies": { "base64-js": "^1.0.2", "ieee754": "^1.1.4", "isarray": "^1.0.0" } }, "sha512-xq+q3SRMOxGivLhBNaUdC64hDTQwejJ+H0T/NB1XMtTVEwNTrfFF3gAxiyW0Bu/xWEGhjVKgUcMhCrUy2+uCWg=="],
 
-    "bun-types": ["bun-types@1.2.16", "", { "dependencies": { "@types/node": "*" } }, "sha512-ciXLrHV4PXax9vHvUrkvun9VPVGOVwbbbBF/Ev1cXz12lyEZMoJpIJABOfPcN9gDJRaiKF9MVbSygLg4NXu3/A=="],
+    "bun-types": ["bun-types@1.3.0", "", { "dependencies": { "@types/node": "*" }, "peerDependencies": { "@types/react": "^19" } }, "sha512-u8X0thhx+yJ0KmkxuEo9HAtdfgCBaM/aI9K90VQcQioAmkVp3SG3FkwWGibUFz3WdXAdcsqOcbU40lK7tbHdkQ=="],
 
     "bytes": ["bytes@3.1.2", "", {}, "sha512-/Nf7TyzTx6S3yRJObOAV7956r8cr2+Oj8AC5dt8wSP3BQAoeX58NoHyCU8P8zGkNXStjTSi6fzO6F0pBdcYbEg=="],
 
@@ -325,8 +325,6 @@
 
     "http-errors/statuses": ["statuses@2.0.1", "", {}, "sha512-RwNA9Z/7PrK06rYLIzFMlaF+l73iwpzsqRIFgbMLbTcLD6cOao82TaWefPXQvB2fOC4AjuYSEndS7N/mTCbkdQ=="],
 
-    "models.dev/@types/bun": ["@types/bun@1.3.0", "", { "dependencies": { "bun-types": "1.3.0" } }, "sha512-+lAGCYjXjip2qY375xX/scJeVRmZ5cY0wyHYyCYxNcdEXrQ4AOe3gACgd4iQ8ksOslJtW4VNxBJ8llUwc3a6AA=="],
-
     "opencontrol/@tsconfig/bun": ["@tsconfig/bun@1.0.7", "", {}, "sha512-udGrGJBNQdXGVulehc1aWT73wkR9wdaGBtB6yL70RJsqwW/yJhIg6ZbRlPOfIUiFNrnBuYLBi9CSmMKfDC7dvA=="],
 
     "opencontrol/hono": ["hono@4.7.4", "", {}, "sha512-Pst8FuGqz3L7tFF+u9Pu70eI0xa5S3LPUmrNd5Jm8nTHze9FxLTK9Kaj5g/k4UcwuJSXTP65SyHOPLrffpcAJg=="],
@@ -334,11 +332,5 @@
     "openid-client/jose": ["jose@4.15.9", "", {}, "sha512-1vUQX+IdDMVPj4k8kOxgUqlcK518yluMuGZwqlr44FS1ppZB/5GWh4rZG89erpOBOJjU/OBsnCVFfapsRz6nEA=="],
 
     "bun-types/@types/node/undici-types": ["undici-types@7.8.0", "", {}, "sha512-9UJ2xGDvQ43tYyVMpuHlsgApydB8ZKfVYTsLDhXkFL/6gfkp+U8xTGdh8pMJv1SpZna0zxG1DwsKZsreLbXBxw=="],
-
-    "models.dev/@types/bun/bun-types": ["bun-types@1.3.0", "", { "dependencies": { "@types/node": "*" }, "peerDependencies": { "@types/react": "^19" } }, "sha512-u8X0thhx+yJ0KmkxuEo9HAtdfgCBaM/aI9K90VQcQioAmkVp3SG3FkwWGibUFz3WdXAdcsqOcbU40lK7tbHdkQ=="],
-
-    "models.dev/@types/bun/bun-types/@types/node": ["@types/node@24.0.3", "", { "dependencies": { "undici-types": "~7.8.0" } }, "sha512-R4I/kzCYAdRLzfiCabn9hxWfbuHS573x+r0dJMkkzThEa7pbrcDWK+9zu3e7aBOouf+rQAciqPFMnxwr0aWgKg=="],
-
-    "models.dev/@types/bun/bun-types/@types/node/undici-types": ["undici-types@7.8.0", "", {}, "sha512-9UJ2xGDvQ43tYyVMpuHlsgApydB8ZKfVYTsLDhXkFL/6gfkp+U8xTGdh8pMJv1SpZna0zxG1DwsKZsreLbXBxw=="],
   }
 }

package.json

@@ -17,13 +17,15 @@
   "scripts": {
     "validate": "bun ./packages/core/script/validate.ts",
     "compare:migrations": "bun ./packages/core/script/compare-model-migrations.ts",
+    "chutes:generate": "bun ./packages/core/script/generate-chutes.ts",
     "helicone:generate": "bun ./packages/core/script/generate-helicone.ts",
     "venice:generate": "bun ./packages/core/script/generate-venice.ts",
     "vercel:generate": "bun ./packages/core/script/generate-vercel.ts",
-    "wandb:generate": "bun ./packages/core/script/generate-wandb.ts"
+    "wandb:generate": "bun ./packages/core/script/generate-wandb.ts",
+    "digitalocean:generate": "bun ./packages/core/script/generate-digitalocean.ts"
   },
   "dependencies": {
-    "@cloudflare/workers-types": "^4.20250801.0",
+    "@cloudflare/workers-types": "^4.20260424.1",
     "sst": "3.17.23"
   }
 }

packages/core/script/generate-chutes.ts

@@ -0,0 +1,589 @@
+#!/usr/bin/env bun
+
+/**
+ * Generates Chutes model TOML files from the Chutes LLM API.
+ *
+ * Flags:
+ * --dry-run: Preview changes without writing files
+ * --new-only: Only create new models, skip updating existing ones
+ * --keep-orphans: Don't delete TOML files for models no longer in the API
+ */
+
+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://llm.chutes.ai/v1/models";
+
+enum SkipZeroFields {
+  LimitContext = "limit.context",
+  LimitOutput = "limit.output",
+}
+
+const Pricing = z.object({
+  prompt: z.number().optional(),
+  completion: z.number().optional(),
+  input_cache_read: z.number().optional(),
+}).passthrough();
+
+const ChutesModel = z.object({
+  id: z.string(),
+  created: z.number(),
+  pricing: Pricing.optional(),
+  context_length: z.number().optional(),
+  max_output_length: z.number().optional(),
+  max_model_len: z.number().optional(),
+  input_modalities: z.array(z.string()).optional(),
+  output_modalities: z.array(z.string()).optional(),
+  supported_features: z.array(z.string()).optional(),
+  supported_sampling_parameters: z.array(z.string()).optional(),
+  quantization: z.string().optional(),
+}).passthrough();
+
+const ChutesResponse = z.object({
+  data: z.array(ChutesModel),
+}).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;
+  };
+  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;
+  };
+  limit: {
+    context: number;
+    output: number;
+  };
+  modalities: {
+    input: string[];
+    output: string[];
+  };
+}
+
+interface Changes {
+  field: string;
+  oldValue: string;
+  newValue: string;
+}
+
+// ── Utility functions ────────────────────────────────────────────────
+
+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);
+}
+
+function formatNumber(n: number): string {
+  if (n >= 1000) {
+    return n.toString().replace(/\B(?=(\d{3})+(?!\d))/g, "_");
+  }
+  return n.toString();
+}
+
+/**
+ * Humanize a model ID into a readable name.
+ * Strips the org prefix and replaces hyphens with spaces.
+ * e.g. "Qwen/Qwen3-32B-TEE" → "Qwen3 32B TEE"
+ */
+function humanizeModelName(modelId: string): string {
+  const parts = modelId.split("/");
+  const modelPart = parts[parts.length - 1];
+  return modelPart.replace(/-/g, " ");
+}
+
+// ── Family inference (same approach as generate-vercel.ts) ───────────
+
+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;
+}
+
+// ── Load existing TOML ───────────────────────────────────────────────
+
+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;
+  }
+}
+
+// ── Merge API data with existing TOML ────────────────────────────────
+
+function mergeModel(
+  apiModel: z.infer<typeof ChutesModel>,
+  existing: ExistingModel | null,
+): MergedModel {
+  const features = new Set(apiModel.supported_features ?? []);
+  const samplingParams = new Set(apiModel.supported_sampling_parameters ?? []);
+  const inputMods = apiModel.input_modalities ?? ["text"];
+  const outputMods = apiModel.output_modalities ?? ["text"];
+
+  // Capabilities from API features
+  const hasAttachment = inputMods.some((m) =>
+    m === "image" || m === "video" || m === "pdf",
+  );
+  const hasReasoning = features.has("reasoning");
+  const hasToolCall = features.has("tools");
+  const hasStructuredOutput = features.has("structured_outputs");
+  const hasTemperature = samplingParams.size > 0
+    ? samplingParams.has("temperature")
+    : true; // default true if no sampling params info
+
+  // Preserve existing values when available (manually specified)
+  const modelName = existing?.name ?? humanizeModelName(apiModel.id);
+  const family = existing?.family ?? inferFamily(apiModel.id, modelName);
+  const knowledge = existing?.knowledge;
+  const interleaved = existing?.interleaved;
+  const status = existing?.status;
+
+  // Release date: existing > API created timestamp > today
+  const releaseDate = existing?.release_date
+    ?? timestampToDate(apiModel.created)
+    ?? getTodayDate();
+
+  // Context limit: prefer context_length, fallback to max_model_len
+  const apiContext = apiModel.context_length ?? apiModel.max_model_len ?? 0;
+  const contextLimit = apiContext > 0
+    ? apiContext
+    : (existing?.limit?.context ?? 0);
+
+  // Output limit: prefer max_output_length, fallback to existing
+  const apiOutput = apiModel.max_output_length ?? 0;
+  const outputLimit = apiOutput > 0
+    ? apiOutput
+    : (existing?.limit?.output ?? 0);
+
+  const merged: MergedModel = {
+    name: modelName,
+    family,
+    attachment: hasAttachment,
+    reasoning: hasReasoning,
+    tool_call: hasToolCall,
+    temperature: hasTemperature,
+    release_date: releaseDate,
+    last_updated: getTodayDate(),
+    open_weights: true, // Chutes hosts open-weight models
+    ...(hasStructuredOutput && { structured_output: hasStructuredOutput }),
+    ...(knowledge && { knowledge }),
+    ...(interleaved !== undefined && { interleaved }),
+    ...(status && { status }),
+    limit: {
+      context: contextLimit,
+      output: outputLimit,
+    },
+    modalities: {
+      input: inputMods,
+      output: outputMods,
+    },
+  };
+
+  // Cost: API values are already in USD per 1M tokens — use directly
+  if (apiModel.pricing) {
+    const inputPrice = apiModel.pricing.prompt;
+    const outputPrice = apiModel.pricing.completion;
+    const cacheReadPrice = apiModel.pricing.input_cache_read;
+
+    if (inputPrice !== undefined && outputPrice !== undefined) {
+      merged.cost = {
+        input: inputPrice,
+        output: outputPrice,
+        ...(cacheReadPrice !== undefined && { cache_read: cacheReadPrice }),
+      };
+    }
+  }
+
+  return merged;
+}
+
+// ── TOML formatting ──────────────────────────────────────────────────
+
+function formatToml(model: MergedModel): string {
+  const lines: string[] = [];
+
+  lines.push(`# Auto-generated by generate-chutes.ts — do not edit pricing, limits, or capabilities.`);
+  lines.push(`# Manual overrides preserved on re-run: name, family, knowledge, interleaved, status`);
+  lines.push(`name = "${model.name.replace(/"/g, '\\"')}"`);
+  if (model.family) {
+    lines.push(`family = "${model.family}"`);
+  }
+  lines.push(`release_date = "${model.release_date}"`);
+  lines.push(`last_updated = "${model.last_updated}"`);
+  lines.push(`attachment = ${model.attachment}`);
+  lines.push(`reasoning = ${model.reasoning}`);
+  lines.push(`temperature = ${model.temperature}`);
+  lines.push(`tool_call = ${model.tool_call}`);
+  if (model.structured_output !== undefined) {
+    lines.push(`structured_output = ${model.structured_output}`);
+  }
+  lines.push(`open_weights = ${model.open_weights}`);
+  if (model.knowledge) {
+    lines.push(`knowledge = "${model.knowledge}"`);
+  }
+  if (model.status) {
+    lines.push(`status = "${model.status}"`);
+  }
+
+  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}`);
+    }
+  }
+
+  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) => `"${m}"`).join(", ")}]`);
+
+  if (model.interleaved !== undefined) {
+    lines.push("");
+    if (model.interleaved === true) {
+      lines.push(`interleaved = true`);
+    } else if (typeof model.interleaved === "object") {
+      lines.push(`[interleaved]`);
+      lines.push(`field = "${model.interleaved.field}"`);
+    }
+  }
+
+  return lines.join("\n") + "\n";
+}
+
+// ── Change detection ─────────────────────────────────────────────────
+
+function detectChanges(
+  existing: ExistingModel | null,
+  merged: MergedModel,
+): Changes[] {
+  if (!existing) return [];
+
+  const changes: Changes[] = [];
+  const EPSILON = 0.001;
+
+  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 => {
+    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("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);
+  compare("modalities.output", existing.modalities?.output, merged.modalities.output);
+
+  return changes;
+}
+
+// ── Main ─────────────────────────────────────────────────────────────
+
+async function main() {
+  const args = process.argv.slice(2);
+  const dryRun = args.includes("--dry-run");
+  const newOnly = args.includes("--new-only");
+  const keepOrphans = args.includes("--keep-orphans");
+
+  const modelsDir = path.join(
+    import.meta.dirname,
+    "..",
+    "..",
+    "..",
+    "providers",
+    "chutes",
+    "models",
+  );
+
+  console.log(`${dryRun ? "[DRY RUN] " : ""}${newOnly ? "[NEW ONLY] " : ""}${keepOrphans ? "[KEEP ORPHANS] " : ""}Fetching Chutes 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 = ChutesResponse.safeParse(json);
+  if (!parsed.success) {
+    console.error("Invalid API response:", parsed.error.errors);
+    process.exit(1);
+  }
+
+  const apiModels = parsed.data.data;
+
+  // Scan existing TOML files
+  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, ${existingFiles.size} existing files\n`);
+
+  const apiModelIds = new Set<string>();
+
+  let created = 0;
+  let updated = 0;
+  let unchanged = 0;
+
+  for (const apiModel of apiModels) {
+    const relativePath = `${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);
+      const existingContent = await Bun.file(filePath).text();
+      const formatChanged = existingContent !== tomlContent;
+
+      if (changes.length > 0 || formatChanged) {
+        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}`);
+        }
+        if (changes.length === 0 && formatChanged) {
+          console.log(`  (format-only change)`);
+        }
+        console.log("");
+      } else {
+        unchanged++;
+      }
+    }
+  }
+
+  // Handle orphaned files (on disk but not in API)
+  const orphaned: string[] = [];
+  for (const file of existingFiles) {
+    if (!apiModelIds.has(file)) {
+      orphaned.push(file);
+      const orphanPath = path.join(modelsDir, file);
+      if (keepOrphans) {
+        console.log(`Orphaned (kept): ${file}`);
+      } else if (dryRun) {
+        console.log(`[DRY RUN] Would delete: ${file}`);
+      } else {
+        await Bun.file(orphanPath).delete();
+        console.log(`Deleted: ${file}`);
+
+        // Clean up empty parent directories
+        const parentDir = path.dirname(orphanPath);
+        try {
+          const remaining = [];
+          for await (const entry of new Bun.Glob("*").scan({ cwd: parentDir })) {
+            remaining.push(entry);
+          }
+          if (remaining.length === 0) {
+            const { rmdir } = await import("node:fs/promises");
+            await rmdir(parentDir);
+            console.log(`  Removed empty directory: ${path.basename(parentDir)}/`);
+          }
+        } catch {
+          // Directory not empty or other error, ignore
+        }
+      }
+    }
+  }
+
+  console.log("");
+  if (dryRun) {
+    console.log(
+      `Summary: ${created} would be created, ${updated} would be updated, ${unchanged} unchanged, ${orphaned.length} would be deleted`,
+    );
+  } else if (keepOrphans) {
+    console.log(
+      `Summary: ${created} created, ${updated} updated, ${unchanged} unchanged, ${orphaned.length} orphaned (kept)`,
+    );
+  } else {
+    console.log(
+      `Summary: ${created} created, ${updated} updated, ${unchanged} unchanged, ${orphaned.length} deleted`,
+    );
+  }
+}
+
+await main();

packages/core/script/generate-digitalocean.ts

@@ -0,0 +1,732 @@
+#!/usr/bin/env bun
+
+/**
+ * Generates DigitalOcean model TOML files from two public APIs:
+ *
+ *   - https://api.digitalocean.com/v2/gen-ai/models   (model metadata, lifecycle, modalities, limits)
+ *   - https://www.digitalocean.com/api/static-content/v1/products  (pricing, including >200k tiers)
+ *
+ * The v2 models API requires a DigitalOcean personal access token or model access key,
+ * read from the DIGITALOCEAN_API_TOKEN environment variable (or --api-key flag).
+ * The static-content pricing API is public and requires no auth.
+ *
+ * Cache pricing (cache_read, cache_write) is NOT available from any DO API and is
+ * preserved from existing TOML files when present.
+ *
+ * Fields the APIs cannot provide (preserved from existing TOMLs, never overwritten):
+ *   family, knowledge, open_weights, interleaved, attachment, release_date,
+ *   cache_read, cache_write
+ *
+ * Flags:
+ *   --dry-run   Preview changes without writing files
+ *   --new-only  Only create new models, skip updating existing ones
+ *   --api-key=<key>  DigitalOcean API key (overrides DIGITALOCEAN_API_TOKEN env var)
+ */
+
+import { z } from "zod";
+import path from "node:path";
+import { mkdir } from "node:fs/promises";
+import { ModelFamilyValues } from "../src/family.js";
+
+const MODELS_API = "https://api.digitalocean.com/v2/gen-ai/models";
+const PRICING_API = "https://www.digitalocean.com/api/static-content/v1/products";
+
+// ---------------------------------------------------------------------------
+// v2 models API schema
+// ---------------------------------------------------------------------------
+
+const DoModel = z
+  .object({
+    id: z.string(),
+    name: z.string(),
+    lifecycle_status: z.string(),
+    type: z.string().optional(),
+    thinking: z.boolean().optional(),
+    context_window: z.union([z.number(), z.string()]).optional(),
+    modalities: z
+      .object({
+        input: z.array(z.string()).optional(),
+        output: z.array(z.string()).optional(),
+      })
+      .optional(),
+    settings: z
+      .array(
+        z.object({
+          name: z.string(),
+          max: z.number().optional(),
+          default_value: z.number().optional(),
+        }),
+      )
+      .optional(),
+    created_at: z.string().optional(),
+  })
+  .passthrough();
+
+const DoModelsResponse = z
+  .object({
+    models: z.array(DoModel),
+  })
+  .passthrough();
+
+// ---------------------------------------------------------------------------
+// static-content pricing API schema
+// ---------------------------------------------------------------------------
+
+const PricingEntry = z
+  .object({
+    name: z.string(),
+    slug: z.string(),
+    model: z.string(),
+    prompt_tokens: z.string().optional(), // "≤200k" | ">200k" | undefined
+    price: z.object({ rate: z.number() }),
+  })
+  .passthrough();
+
+const StaticContentResponse = z
+  .object({
+    gradient: z.object({
+      models: z.array(PricingEntry),
+    }),
+  })
+  .passthrough();
+
+// ---------------------------------------------------------------------------
+// Derived pricing map
+// ---------------------------------------------------------------------------
+
+interface ModelPricing {
+  input: number;
+  output: number;
+  inputOver200k?: number;
+  outputOver200k?: number;
+}
+
+// Map marketing names from /v1/products to API model IDs from /v2/gen-ai/models.
+// The pricing API uses display names, not the machine IDs, so this table is the
+// join key. Add entries here when DO adds new models with tiered pricing.
+const PRICING_NAME_MAP: Record<string, string> = {
+  // Anthropic
+  "claude sonnet 4.6": "anthropic-claude-4.6-sonnet",
+  "claude sonnet 4.5": "anthropic-claude-4.5-sonnet",
+  "claude sonnet 4": "anthropic-claude-sonnet-4",
+  "claude haiku 4.5": "anthropic-claude-haiku-4.5",
+  "claude opus 4.6": "anthropic-claude-opus-4.6",
+  "claude opus 4.5": "anthropic-claude-opus-4.5",
+  "claude opus 4.1": "anthropic-claude-4.1-opus",
+  "claude opus 4": "anthropic-claude-opus-4",
+  // OpenAI
+  "gpt-5.4": "openai-gpt-5.4",
+  "gpt-5.4 mini": "openai-gpt-5.4-mini",
+  "gpt-5.4 nano": "openai-gpt-5.4-nano",
+  "gpt-5.4 pro": "openai-gpt-5.4-pro",
+  "gpt-5.3-codex": "openai-gpt-5.3-codex",
+  "gpt-5.2": "openai-gpt-5.2",
+  "gpt-5.2 pro": "openai-gpt-5.2-pro",
+  "gpt-5.1-codex-max": "openai-gpt-5.1-codex-max",
+  "gpt-5": "openai-gpt-5",
+  "gpt-5 mini": "openai-gpt-5-mini",
+  "gpt-5 nano": "openai-gpt-5-nano",
+  "gpt-4.1": "openai-gpt-4.1",
+  "gpt image 1": "openai-gpt-image-1",
+  "gpt image 1.5": "openai-gpt-image-1.5",
+  "gpt-oss-120b": "openai-gpt-oss-120b",
+  "gpt-oss-20b": "openai-gpt-oss-20b",
+  "gpt-4o": "openai-gpt-4o",
+  "gpt-4o mini": "openai-gpt-4o-mini",
+  "o1": "openai-o1",
+  "o3-mini": "openai-o3-mini",
+  // DeepSeek
+  "deepseek r1 distill llama 70b": "deepseek-r1-distill-llama-70b",
+  // Llama
+  "llama 3.3 70b": "llama3.3-70b-instruct",
+  // DO-hosted
+  "qwen3-32b": "alibaba-qwen3-32b",
+  "minimax m2.5 (public preview)": "minimax-m2.5",
+  "kimi k2.5": "kimi-k2.5",
+  "nvidia nemotron 3 super 120b (public preview)": "nvidia-nemotron-3-super-120b",
+  "glm 5": "glm-5",
+};
+
+function normalizeDisplayName(raw: string): string {
+  // Strip " Input Tokens" / " Output Tokens" suffix and lowercase
+  return raw
+    .replace(/\s+(input|output)\s+tokens$/i, "")
+    .trim()
+    .toLowerCase();
+}
+
+function buildPricingMap(entries: z.infer<typeof PricingEntry>[]): Map<string, ModelPricing> {
+  const map = new Map<string, ModelPricing>();
+
+  for (const entry of entries) {
+    const displayName = normalizeDisplayName(entry.name);
+    const modelId = PRICING_NAME_MAP[displayName];
+    if (!modelId) continue;
+
+    const isInput = entry.name.toLowerCase().includes("input tokens");
+    const isOver200k = entry.prompt_tokens === ">200k";
+    // Round to avoid float noise (e.g. 0.9900000000000001)
+    const rate = Math.round(entry.price.rate * 10000) / 10000;
+
+    const existing = map.get(modelId) ?? ({} as ModelPricing);
+
+    if (isInput && isOver200k) existing.inputOver200k = rate;
+    else if (!isInput && isOver200k) existing.outputOver200k = rate;
+    else if (isInput) existing.input = rate;
+    else existing.output = rate;
+
+    map.set(modelId, existing);
+  }
+
+  return map;
+}
+
+// ---------------------------------------------------------------------------
+// Existing TOML shape (fields we read and may preserve)
+// ---------------------------------------------------------------------------
+
+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;
+    context_over_200k?: {
+      input?: number;
+      output?: number;
+      cache_read?: number;
+      cache_write?: number;
+      context_min?: number;
+    };
+    tiers?: Array<{
+      tier: {
+        type?: "context";
+        size: number;
+      };
+      input?: number;
+      output?: number;
+      cache_read?: number;
+      cache_write?: number;
+    }>;
+  };
+  limit?: {
+    context?: number;
+    input?: number;
+    output?: number;
+  };
+  modalities?: {
+    input?: string[];
+    output?: string[];
+  };
+}
+
+async function loadExisting(filePath: string): Promise<ExistingModel | null> {
+  const file = Bun.file(filePath);
+  if (!(await file.exists())) return null;
+  try {
+    const mod = await import(filePath, { with: { type: "toml" } });
+    return mod.default as ExistingModel;
+  } catch (e) {
+    console.warn(`Warning: failed to parse ${filePath}:`, e);
+    return null;
+  }
+}
+
+// ---------------------------------------------------------------------------
+// Merged model shape (what we write)
+// ---------------------------------------------------------------------------
+
+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;
+    context_over_200k?: {
+      input: number;
+      output: number;
+      cache_read?: number;
+      cache_write?: number;
+      context_min?: number;
+    };
+  };
+  limit: {
+    context: number;
+    output: number;
+  };
+  modalities: {
+    input: string[];
+    output: string[];
+  };
+}
+
+// ---------------------------------------------------------------------------
+// Helpers
+// ---------------------------------------------------------------------------
+
+const VALID_INPUT_MODALITIES = new Set(["text", "audio", "image", "video", "pdf"]);
+const VALID_OUTPUT_MODALITIES = new Set(["text", "audio", "image", "video", "pdf"]);
+
+function filterInputModalities(raw: string[]): string[] {
+  return raw.filter((m) => VALID_INPUT_MODALITIES.has(m));
+}
+
+function filterOutputModalities(raw: string[]): string[] {
+  // "code" is not a valid modality in the schema — map to "text"
+  return [...new Set(raw.map((m) => (m === "code" ? "text" : m)).filter((m) => VALID_OUTPUT_MODALITIES.has(m)))];
+}
+
+function getTodayDate(): string {
+  return new Date().toISOString().slice(0, 10);
+}
+
+function formatNumber(n: number): string {
+  return n >= 1000 ? n.toString().replace(/\B(?=(\d{3})+(?!\d))/g, "_") : n.toString();
+}
+
+function inferFamily(modelId: string, modelName: string): string | undefined {
+  const sorted = [...ModelFamilyValues].sort((a, b) => b.length - a.length);
+  const targets = [modelId.toLowerCase(), modelName.toLowerCase()];
+  for (const family of sorted) {
+    const f = family.toLowerCase();
+    for (const t of targets) {
+      if (t.includes(f)) return family;
+    }
+  }
+  return undefined;
+}
+
+function getExistingLongContextCost(existing: ExistingModel | null) {
+  const tier = existing?.cost?.tiers?.find(
+    (tier) =>
+      (tier.tier.type === undefined || tier.tier.type === "context") &&
+      tier.tier.size >= 200_000,
+  );
+  if (tier) {
+    return {
+      ...tier,
+      context_min: tier.tier.size,
+    };
+  }
+
+  return existing?.cost?.context_over_200k === undefined
+    ? undefined
+    : {
+        ...existing.cost.context_over_200k,
+        context_min: 200_000,
+      };
+}
+
+function getLongContextMin(cost: { context_min?: number }) {
+  return cost.context_min ?? 200_000;
+}
+
+function formatInlineNumber(n: number): string {
+  return n >= 1000 ? n.toString().replace(/\B(?=(\d{3})+(?!\d))/g, "_") : n.toString();
+}
+
+// ---------------------------------------------------------------------------
+// Merge API data with existing TOML
+// ---------------------------------------------------------------------------
+
+function mergeModel(
+  apiModel: z.infer<typeof DoModel>,
+  pricing: ModelPricing | undefined,
+  existing: ExistingModel | null,
+): MergedModel {
+  const rawInput = apiModel.modalities?.input ?? [];
+  const rawOutput = apiModel.modalities?.output ?? [];
+  const inputMods = filterInputModalities(rawInput.length > 0 ? rawInput : existing?.modalities?.input ?? ["text"]);
+  const outputMods = filterOutputModalities(rawOutput.length > 0 ? rawOutput : existing?.modalities?.output ?? ["text"]);
+
+  const maxTokensSetting = apiModel.settings?.find((s) => s.name === "max_tokens");
+  const maxTokens = maxTokensSetting?.max ?? existing?.limit?.output ?? 0;
+
+  const rawContext = apiModel.context_window;
+  const contextWindow =
+    rawContext !== undefined
+      ? typeof rawContext === "string"
+        ? parseInt(rawContext, 10)
+        : rawContext
+      : (existing?.limit?.context ?? 0);
+
+  const isDeprecated = apiModel.lifecycle_status === "end_of_life";
+
+  // Fields preserved from existing TOML (APIs don't provide these)
+  const family = existing?.family ?? inferFamily(apiModel.id, apiModel.name);
+  const knowledge = existing?.knowledge;
+  const openWeights = existing?.open_weights ?? false;
+  const interleaved = existing?.interleaved;
+  const attachment = existing?.attachment ?? inputMods.some((m) => m !== "text");
+
+  // reasoning: trust existing if set, else use API thinking flag as a hint
+  // (thinking flag is unreliable for non-LLM models so gate on output modality)
+  const isTextOutput = outputMods.includes("text") && !outputMods.includes("image") && !outputMods.includes("video");
+  const reasoning = existing?.reasoning ?? (isTextOutput && (apiModel.thinking ?? false));
+
+  // tool_call: no API signal, preserve existing or default true for text models
+  const toolCall = existing?.tool_call ?? isTextOutput;
+
+  // temperature: no API signal, preserve or default true
+  const temperature = existing?.temperature ?? true;
+
+  // structured_output: no API signal, preserve only
+  const structuredOutput = existing?.structured_output;
+
+  const releaseDate = existing?.release_date ?? apiModel.created_at?.slice(0, 10) ?? getTodayDate();
+
+  const merged: MergedModel = {
+    name: apiModel.name,
+    family,
+    attachment,
+    reasoning,
+    tool_call: toolCall,
+    temperature,
+    release_date: releaseDate,
+    last_updated: getTodayDate(),
+    open_weights: openWeights,
+    ...(structuredOutput !== undefined && { structured_output: structuredOutput }),
+    ...(knowledge && { knowledge }),
+    ...(interleaved !== undefined && { interleaved }),
+    ...(isDeprecated && { status: "deprecated" }),
+    limit: { context: contextWindow, output: maxTokens },
+    modalities: { input: inputMods, output: outputMods },
+  };
+
+  // Pricing: static-content API is the sole source of truth for prices.
+  // The v2 models API pricing is intentionally ignored. If a model has no
+  // entry in the static-content API, preserve existing TOML prices.
+  const inputPrice = pricing?.input ?? existing?.cost?.input;
+  const outputPrice = pricing?.output ?? existing?.cost?.output;
+
+  if (inputPrice !== undefined && outputPrice !== undefined) {
+    merged.cost = {
+      input: inputPrice,
+      output: outputPrice,
+      // Always preserve cache pricing — not available from any DO API
+      ...(existing?.cost?.cache_read !== undefined && { cache_read: existing.cost.cache_read }),
+      ...(existing?.cost?.cache_write !== undefined && { cache_write: existing.cost.cache_write }),
+    };
+
+    // Context-tiered pricing (>200k) from the static-content API
+    const existingLongContextCost = getExistingLongContextCost(existing);
+    if (pricing?.inputOver200k !== undefined && pricing?.outputOver200k !== undefined) {
+      merged.cost.context_over_200k = {
+        input: pricing.inputOver200k,
+        output: pricing.outputOver200k,
+        context_min: existingLongContextCost?.context_min ?? 200_000,
+        ...(existingLongContextCost?.cache_read !== undefined && {
+          cache_read: existingLongContextCost.cache_read,
+        }),
+        ...(existingLongContextCost?.cache_write !== undefined && {
+          cache_write: existingLongContextCost.cache_write,
+        }),
+      };
+    } else if (existingLongContextCost) {
+      // Preserve manually-entered tiered pricing if API has no data
+      merged.cost.context_over_200k = {
+        input: existingLongContextCost.input ?? inputPrice,
+        output: existingLongContextCost.output ?? outputPrice,
+        context_min: existingLongContextCost.context_min,
+        ...(existingLongContextCost.cache_read !== undefined && {
+          cache_read: existingLongContextCost.cache_read,
+        }),
+        ...(existingLongContextCost.cache_write !== undefined && {
+          cache_write: existingLongContextCost.cache_write,
+        }),
+      };
+    }
+  }
+
+  return merged;
+}
+
+// ---------------------------------------------------------------------------
+// TOML serialiser
+// ---------------------------------------------------------------------------
+
+function formatToml(model: MergedModel): string {
+  const lines: string[] = [];
+
+  lines.push(`name = "${model.name.replace(/"/g, '\\"')}"`);
+  if (model.family) lines.push(`family = "${model.family}"`);
+  lines.push(`release_date = "${model.release_date}"`);
+  lines.push(`last_updated = "${model.last_updated}"`);
+  lines.push(`attachment = ${model.attachment}`);
+  lines.push(`reasoning = ${model.reasoning}`);
+  lines.push(`temperature = ${model.temperature}`);
+  lines.push(`tool_call = ${model.tool_call}`);
+  if (model.structured_output !== undefined) lines.push(`structured_output = ${model.structured_output}`);
+  if (model.knowledge) lines.push(`knowledge = "${model.knowledge}"`);
+  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}`);
+
+    if (model.cost.context_over_200k) {
+      lines.push("");
+      lines.push(`[[cost.tiers]]`);
+      lines.push(`tier = { size = ${formatInlineNumber(getLongContextMin(model.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}`);
+    }
+  }
+
+  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) => `"${m}"`).join(", ")}]`);
+
+  return lines.join("\n") + "\n";
+}
+
+// ---------------------------------------------------------------------------
+// Change detection
+// ---------------------------------------------------------------------------
+
+interface Change {
+  field: string;
+  oldValue: string;
+  newValue: string;
+}
+
+function formatValue(val: unknown): string {
+  if (val === undefined) return "(none)";
+  if (Array.isArray(val)) return `[${val.join(", ")}]`;
+  if (typeof val === "number") return formatNumber(val);
+  return String(val);
+}
+
+function detectChanges(existing: ExistingModel | null, merged: MergedModel): Change[] {
+  if (!existing) return [];
+
+  const changes: Change[] = [];
+  const EPSILON = 0.001;
+
+  const compare = (field: string, oldVal: unknown, newVal: unknown) => {
+    if (oldVal === undefined && newVal === undefined) return;
+    const isDiff = field.startsWith("cost.")
+      ? Math.abs((oldVal as number ?? 0) - (newVal as number ?? 0)) > EPSILON
+      : JSON.stringify(oldVal) !== JSON.stringify(newVal);
+    if (isDiff) changes.push({ field, oldValue: formatValue(oldVal), newValue: formatValue(newVal) });
+  };
+
+  compare("name", existing.name, merged.name);
+  compare("reasoning", existing.reasoning, merged.reasoning);
+  compare("tool_call", existing.tool_call, merged.tool_call);
+  compare("attachment", existing.attachment, merged.attachment);
+  compare("status", existing.status, merged.status);
+  compare("cost.input", existing.cost?.input, merged.cost?.input);
+  compare("cost.output", existing.cost?.output, merged.cost?.output);
+  const existingLongContextCost = getExistingLongContextCost(existing);
+  compare("cost.context_over_200k.input", existingLongContextCost?.input, merged.cost?.context_over_200k?.input);
+  compare("cost.context_over_200k.output", existingLongContextCost?.output, merged.cost?.context_over_200k?.output);
+  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);
+  compare("modalities.output", existing.modalities?.output, merged.modalities.output);
+
+  return changes;
+}
+
+// ---------------------------------------------------------------------------
+// Main
+// ---------------------------------------------------------------------------
+
+async function main() {
+  const args = process.argv.slice(2);
+  const dryRun = args.includes("--dry-run");
+  const newOnly = args.includes("--new-only");
+
+  // Resolve API key
+  const apiKeyArg = args.find((a) => a.startsWith("--api-key"));
+  const apiKey =
+    (apiKeyArg?.includes("=") ? apiKeyArg.split("=")[1] : args[args.indexOf(apiKeyArg!) + 1]) ??
+    process.env.DIGITALOCEAN_API_TOKEN;
+
+  if (!apiKey) {
+    console.error("Error: DIGITALOCEAN_API_TOKEN is required (or pass --api-key=<key>)");
+    console.error("Get one from: https://cloud.digitalocean.com/account/api/tokens");
+    process.exit(1);
+  }
+
+  const modelsDir = path.join(import.meta.dirname, "..", "..", "..", "providers", "digitalocean", "models");
+
+  const prefix = dryRun ? "[DRY RUN] " : "";
+  console.log(`${prefix}Fetching DigitalOcean models from API...`);
+
+  // Fetch both APIs in parallel
+  const [modelsRes, pricingRes] = await Promise.all([
+    fetch(MODELS_API, { headers: { Authorization: `Bearer ${apiKey}`, "Content-Type": "application/json" } }),
+    fetch(PRICING_API, { headers: { "User-Agent": "models.dev/digitalocean-sync" } }),
+  ]);
+
+  if (!modelsRes.ok) {
+    console.error(`Failed to fetch models API: ${modelsRes.status} ${modelsRes.statusText}`);
+    if (modelsRes.status === 401 || modelsRes.status === 403)
+      console.error("Check your DIGITALOCEAN_API_TOKEN has read access.");
+    process.exit(1);
+  }
+
+  if (!pricingRes.ok) {
+    console.error(`Failed to fetch pricing API: ${pricingRes.status} ${pricingRes.statusText}`);
+    process.exit(1);
+  }
+
+  const modelsParsed = DoModelsResponse.safeParse(await modelsRes.json());
+  if (!modelsParsed.success) {
+    console.error("Unexpected models API response:", modelsParsed.error.errors);
+    process.exit(1);
+  }
+
+  const pricingParsed = StaticContentResponse.safeParse(await pricingRes.json());
+  if (!pricingParsed.success) {
+    console.error("Unexpected pricing API response:", pricingParsed.error.errors);
+    process.exit(1);
+  }
+
+  const apiModels = modelsParsed.data.models;
+  const pricingMap = buildPricingMap(pricingParsed.data.gradient.models);
+
+  // Collect existing TOML filenames for orphan detection
+  const existingFiles = new Set<string>();
+  for await (const file of new Bun.Glob("**/*.toml").scan({ cwd: modelsDir, absolute: false })) {
+    existingFiles.add(file);
+  }
+
+  console.log(`Found ${apiModels.length} models in API, ${existingFiles.size} existing TOML files\n`);
+
+  const apiModelFiles = new Set<string>();
+  let created = 0;
+  let updated = 0;
+  let unchanged = 0;
+
+  for (const apiModel of apiModels) {
+    // Skip non-text models that opencode can't use: image, video, audio, embedding, reranking
+    const outputMods = filterOutputModalities(apiModel.modalities?.output ?? []);
+    const isTextModel = outputMods.includes("text");
+    const isEmbedding = apiModel.type === "embedding";
+    const isReranking = apiModel.type === "reranking";
+    if (!isTextModel || isEmbedding || isReranking) continue;
+
+    // Model IDs may contain slashes (e.g. fal-ai/flux/schnell) — use as subpath
+    const relativePath = `${apiModel.id}.toml`;
+    const filePath = path.join(modelsDir, relativePath);
+    const dirPath = path.dirname(filePath);
+
+    apiModelFiles.add(relativePath);
+
+    const existing = await loadExisting(filePath);
+    const pricing = pricingMap.get(apiModel.id);
+    const merged = mergeModel(apiModel, pricing, existing);
+    const toml = formatToml(merged);
+
+    if (existing === null) {
+      created++;
+      if (dryRun) {
+        console.log(`[DRY RUN] Would create: ${relativePath}`);
+        console.log(`  name = "${merged.name}"`);
+        if (pricing) console.log(`  pricing: $${merged.cost?.input}/$${merged.cost?.output} per M tokens`);
+        if (merged.family) console.log(`  family = "${merged.family}" (inferred)`);
+        console.log("");
+      } else {
+        await mkdir(dirPath, { recursive: true });
+        await Bun.write(filePath, toml);
+        console.log(`Created: ${relativePath}`);
+      }
+      continue;
+    }
+
+    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, toml);
+        console.log(`Updated: ${relativePath}`);
+      }
+      for (const c of changes) console.log(`  ${c.field}: ${c.oldValue} → ${c.newValue}`);
+      console.log("");
+    } else {
+      unchanged++;
+    }
+  }
+
+  // Orphan detection: files in the TOML directory but not in the API
+  const orphaned: string[] = [];
+  for (const file of existingFiles) {
+    if (!apiModelFiles.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/script/generate-venice.ts

@@ -162,7 +162,18 @@ interface ExistingModel {
       output?: number;
       cache_read?: number;
       cache_write?: number;
+      context_min?: number;
     };
+    tiers?: Array<{
+      tier: {
+        type?: "context";
+        size: number;
+      };
+      input?: number;
+      output?: number;
+      cache_read?: number;
+      cache_write?: number;
+    }>;
   };
   limit?: {
     context?: number;
@@ -195,6 +206,30 @@ async function loadExistingModel(filePath: string): Promise<ExistingModel | null
   }
 }
 
+function getExistingLongContextMin(existing: ExistingModel | null) {
+  return (
+    existing?.cost?.tiers?.find(
+      (tier) =>
+        (tier.tier.type === undefined || tier.tier.type === "context") &&
+        tier.tier.size >= 200_000,
+    )?.tier.size ?? 200_000
+  );
+}
+
+function getExistingLongContextCost(existing: ExistingModel | null) {
+  return (
+    existing?.cost?.tiers?.find(
+      (tier) =>
+        (tier.tier.type === undefined || tier.tier.type === "context") &&
+        tier.tier.size >= 200_000,
+    ) ?? existing?.cost?.context_over_200k
+  );
+}
+
+function getLongContextMin(cost: { context_min?: number }) {
+  return cost.context_min ?? 200_000;
+}
+
 interface MergedModel {
   name: string;
   family?: string;
@@ -219,6 +254,7 @@ interface MergedModel {
       output: number;
       cache_read?: number;
       cache_write?: number;
+      context_min?: number;
     };
   };
   limit: {
@@ -292,6 +328,7 @@ function mergeModel(
       merged.cost.context_over_200k = {
         input: spec.pricing.extended.input.usd,
         output: spec.pricing.extended.output.usd,
+        context_min: spec.pricing.extended.context_token_threshold,
         ...(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 }),
       };
@@ -366,7 +403,8 @@ function formatToml(model: MergedModel): string {
 
     if (model.cost.context_over_200k) {
       lines.push("");
-      lines.push(`[cost.context_over_200k]`);
+      lines.push(`[[cost.tiers]]`);
+      lines.push(`tier = { size = ${formatNumber(getLongContextMin(model.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) {
@@ -438,10 +476,11 @@ 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);
+  const existingLongContextCost = getExistingLongContextCost(existing);
+  compare("cost.context_over_200k.input", existingLongContextCost?.input, merged.cost?.context_over_200k?.input);
+  compare("cost.context_over_200k.output", existingLongContextCost?.output, merged.cost?.context_over_200k?.output);
+  compare("cost.context_over_200k.cache_read", existingLongContextCost?.cache_read, merged.cost?.context_over_200k?.cache_read);
+  compare("cost.context_over_200k.cache_write", existingLongContextCost?.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);

packages/core/src/family.ts

@@ -54,12 +54,17 @@ export const ModelFamilyValues = [
   // DeepSeek
   "deepseek",
   "deepseek-thinking",
+  "deepseek-flash",
+  "deepseek-flash-free",
+  "deepseek-flash-think",
 
   // Microsoft Phi
   "phi",
 
   // Moonshot Kimi
   "kimi",
+  "kimi-k2.5",
+  "kimi-k2.6",
   "kimi-free",
   "kimi-thinking",
 
@@ -115,8 +120,8 @@ export const ModelFamilyValues = [
   // Hunyuan
   "hunyuan",
 
-  // HY
-  "HY",
+  // Hy
+  "Hy",
 
   // Yi
   "yi",
@@ -205,6 +210,10 @@ export const ModelFamilyValues = [
   "mimo",
   "mimo-pro",
   "mimo-omni",
+  "mimo-v2-pro",
+  "mimo-v2-omni",
+  "mimo-v2.5-pro",
+  "mimo-v2.5",
   "mimo-pro-free",
   "mimo-omni-free",
   "mimo-flash-free",
@@ -280,9 +289,15 @@ export const ModelFamilyValues = [
   // RNJ
   "rnj",
 
+  // Tecent Hy
+  "hy3",
+  "hy3-free",
+
   // Ling & Ring (InclusionAI)
   "ling",
+  "ling-flash-free",
   "ring",
+  "ring-1t-free",
 
   // Kat Coder
   "kat-coder",

packages/core/src/generate.ts

@@ -2,9 +2,9 @@ import path from "path";
 import { mergeDeep } from "remeda";
 import { z } from "zod";
 
-import { Provider, Model } from "./schema.js";
+import { Provider, Model, AuthoredModel, AuthoredModelShape } from "./schema.js";
 
-const ExtendsModel = Model.sourceType()
+const ExtendsModel = AuthoredModelShape
   .partial()
   .extend({
     extends: z
@@ -71,12 +71,12 @@ export async function generate(directory: string) {
         });
         continue;
       }
-      const model = Model.safeParse(toml);
+      const model = AuthoredModel.safeParse(toml);
       if (!model.success) {
         model.error.cause = { modelPath, toml };
         throw model.error;
       }
-      provider.data.models[modelID] = model.data;
+      provider.data.models[modelID] = normalizeModelCost(model.data);
     }
     result[providerID] = provider.data;
   }
@@ -144,7 +144,7 @@ export async function generate(directory: string) {
       }
     }
 
-    const model = Model.safeParse(merged);
+    const model = Model.safeParse(normalizeCost(merged));
     if (!model.success) {
       model.error.cause = { modelPath: pendingModel.modelPath, toml: merged };
       throw model.error;
@@ -155,3 +155,51 @@ export async function generate(directory: string) {
 
   return result;
 }
+
+function normalizeModelCost(model: z.infer<typeof AuthoredModel>): Model {
+  return normalizeCost(model) as Model;
+}
+
+function normalizeCost(model: Record<string, unknown>) {
+  const cost = model.cost;
+  if (cost === undefined || cost === null || typeof cost !== "object" || Array.isArray(cost)) {
+    return model;
+  }
+
+  const tiers = (cost as { tiers?: unknown }).tiers;
+  if (!Array.isArray(tiers)) {
+    return model;
+  }
+
+  if (tiers.length !== 1) {
+    return model;
+  }
+
+  const contextOver200k = tiers.find((tier) => {
+    if (tier === null || typeof tier !== "object" || Array.isArray(tier)) return false;
+    const tierConfig = (tier as { tier?: unknown }).tier;
+    if (tierConfig === null || typeof tierConfig !== "object" || Array.isArray(tierConfig)) return false;
+    const type = (tierConfig as { type?: unknown }).type;
+    const size = (tierConfig as { size?: unknown }).size;
+    // context_over_200k is a legacy compatibility field. It intentionally
+    // includes higher thresholds; cost.tiers carries the exact threshold.
+    return (
+      (type === undefined || type === "context") &&
+      typeof size === "number" &&
+      size >= 200_000
+    );
+  });
+
+  if (contextOver200k === undefined) {
+    return model;
+  }
+
+  const { tier: _tier, ...legacyCost } = contextOver200k as Record<string, unknown>;
+  return {
+    ...model,
+    cost: {
+      ...(cost as Record<string, unknown>),
+      context_over_200k: legacyCost,
+    },
+  };
+}

packages/core/src/schema.ts

@@ -21,110 +21,164 @@ const JsonValue: z.ZodType<JsonValue> = z.lazy(() =>
   ]),
 );
 
-const Cost = z.object({
-  input: z.number().min(0, "Input price cannot be negative"),
-  output: z.number().min(0, "Output price cannot be negative"),
-  reasoning: z.number().min(0, "Input price cannot be negative").optional(),
-  cache_read: z
-    .number()
-    .min(0, "Cache read price cannot be negative")
+const Cost = z
+  .object({
+    input: z.number().min(0, "Input price cannot be negative"),
+    output: z.number().min(0, "Output price cannot be negative"),
+    reasoning: z
+      .number()
+      .min(0, "Reasoning price cannot be negative")
+      .optional(),
+    cache_read: z
+      .number()
+      .min(0, "Cache read price cannot be negative")
+      .optional(),
+    cache_write: z
+      .number()
+      .min(0, "Cache write price cannot be negative")
+      .optional(),
+    input_audio: z
+      .number()
+      .min(0, "Audio input price cannot be negative")
+      .optional(),
+    output_audio: z
+      .number()
+      .min(0, "Audio output price cannot be negative")
+      .optional(),
+  });
+
+const CostTier = Cost.extend({
+  tier: z
+    .object({
+      type: z.literal("context").default("context"),
+      size: z.number().int().min(0, "Context tier size cannot be negative"),
+    })
+    .strict(),
+}).strict();
+
+const AuthoredCost = Cost.extend({
+  context_over_200k: z.never().optional(),
+  tiers: z.array(CostTier).optional(),
+});
+
+const OutputCost = Cost.extend({
+  context_over_200k: Cost.optional(),
+  tiers: z.array(CostTier).optional(),
+});
+
+const ModelBase = z.object({
+  id: z.string(),
+  name: z.string().min(1, "Model name cannot be empty"),
+  family: ModelFamily.optional(),
+  attachment: z.boolean(),
+  reasoning: z.boolean(),
+  tool_call: z.boolean(),
+  interleaved: z
+    .union([
+      z.literal(true),
+      z
+        .object({
+          field: z.enum(["reasoning_content", "reasoning_details"]),
+        })
+        .strict(),
+    ])
     .optional(),
-  cache_write: z
-    .number()
-    .min(0, "Cache write price cannot be negative")
+  structured_output: z.boolean().optional(),
+  temperature: z.boolean().optional(),
+  knowledge: z
+    .string()
+    .regex(/^\d{4}-\d{2}(-\d{2})?$/, {
+      message: "Must be in YYYY-MM or YYYY-MM-DD format",
+    })
     .optional(),
-  input_audio: z
-    .number()
-    .min(0, "Audio input price cannot be negative")
+  release_date: z.string().regex(/^\d{4}-\d{2}(-\d{2})?$/, {
+    message: "Must be in YYYY-MM or YYYY-MM-DD format",
+  }),
+  last_updated: z.string().regex(/^\d{4}-\d{2}(-\d{2})?$/, {
+    message: "Must be in YYYY-MM or YYYY-MM-DD format",
+  }),
+  modalities: z.object({
+    input: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
+    output: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
+  }),
+  open_weights: z.boolean(),
+  limit: z.object({
+    context: z.number().min(0, "Context window must be positive"),
+    input: z.number().min(0, "Input tokens must be positive").optional(),
+    output: z.number().min(0, "Output tokens must be positive"),
+  }),
+  status: z.enum(["alpha", "beta", "deprecated"]).optional(),
+  experimental: z
+    .object({
+      modes: z
+        .record(
+          z.object({
+            cost: Cost.optional(),
+            provider: z
+              .object({
+                body: z.record(JsonValue).optional(),
+                headers: z.record(z.string()).optional(),
+              })
+              .optional(),
+          }),
+        )
+        .optional(),
+    })
     .optional(),
-  output_audio: z
-    .number()
-    .min(0, "Audio output price cannot be negative")
+  provider: z
+    .object({
+      npm: z.string().optional(),
+      api: z.string().optional(),
+      shape: z.enum(["responses", "completions"]).optional(),
+      body: z.record(JsonValue).optional(),
+      headers: z.record(z.string()).optional(),
+    })
     .optional(),
 });
-export const Model = z
+
+function refineModel<T extends z.ZodTypeAny>(schema: T) {
+  return schema
+    .refine(
+      (data) => {
+        return !(data.reasoning === false && data.cost?.reasoning !== undefined);
+      },
+      {
+        message: "Cannot set cost.reasoning when reasoning is false",
+        path: ["cost", "reasoning"],
+      },
+    )
+    .refine(
+      (data) => {
+        const tiers = data.cost?.tiers;
+        if (tiers === undefined) return true;
+
+        const sizes = tiers.map((tier: { tier: { size: number } }) => tier.tier.size);
+        return new Set(sizes).size === sizes.length;
+      },
+      {
+        message: "Cost context tiers must not have duplicate sizes",
+        path: ["cost", "tiers"],
+      },
+    );
+}
+
+export const ModelShape = z
   .object({
-    id: z.string(),
-    name: z.string().min(1, "Model name cannot be empty"),
-    family: ModelFamily.optional(),
-    attachment: z.boolean(),
-    reasoning: z.boolean(),
-    tool_call: z.boolean(),
-    interleaved: z
-      .union([
-        z.literal(true),
-        z
-          .object({
-            field: z.enum(["reasoning_content", "reasoning_details"]),
-          })
-          .strict(),
-      ])
-      .optional(),
-    structured_output: z.boolean().optional(),
-    temperature: z.boolean().optional(),
-    knowledge: z
-      .string()
-      .regex(/^\d{4}-\d{2}(-\d{2})?$/, {
-        message: "Must be in YYYY-MM or YYYY-MM-DD format",
-      })
-      .optional(),
-    release_date: z.string().regex(/^\d{4}-\d{2}(-\d{2})?$/, {
-      message: "Must be in YYYY-MM or YYYY-MM-DD format",
-    }),
-    last_updated: z.string().regex(/^\d{4}-\d{2}(-\d{2})?$/, {
-      message: "Must be in YYYY-MM or YYYY-MM-DD format",
-    }),
-    modalities: z.object({
-      input: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
-      output: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
-    }),
-    open_weights: z.boolean(),
-    cost: Cost.extend({
-      context_over_200k: Cost.optional(),
-    }).optional(),
-    limit: z.object({
-      context: z.number().min(0, "Context window must be positive"),
-      input: z.number().min(0, "Input tokens must be positive").optional(),
-      output: z.number().min(0, "Output tokens must be positive"),
-    }),
-    status: z.enum(["alpha", "beta", "deprecated"]).optional(),
-    experimental: z
-      .object({
-        modes: z
-          .record(
-            z.object({
-              cost: Cost.optional(),
-              provider: z
-                .object({
-                  body: z.record(JsonValue).optional(),
-                  headers: z.record(z.string()).optional(),
-                })
-                .optional(),
-            }),
-          )
-          .optional(),
-      })
-      .optional(),
-    provider: z
-      .object({
-        npm: z.string().optional(),
-        api: z.string().optional(),
-        shape: z.enum(["responses", "completions"]).optional(),
-        body: z.record(JsonValue).optional(),
-        headers: z.record(z.string()).optional(),
-      })
-      .optional(),
+    ...ModelBase.shape,
+    cost: OutputCost.optional(),
   })
-  .strict()
-  .refine(
-    (data) => {
-      return !(data.reasoning === false && data.cost?.reasoning !== undefined);
-    },
-    {
-      message: "Cannot set cost.reasoning when reasoning is false",
-      path: ["cost", "reasoning"],
-    },
-  );
+  .strict();
+
+export const AuthoredModelShape = z
+  .object({
+    ...ModelBase.shape,
+    cost: AuthoredCost.optional(),
+  })
+  .strict();
+
+export const Model = refineModel(ModelShape);
+
+export const AuthoredModel = refineModel(AuthoredModelShape);
 
 export type Model = z.infer<typeof Model>;
 
@@ -150,6 +204,7 @@ export const Provider = z
       const isOpenAIcompatible = data.npm === "@ai-sdk/openai-compatible";
       const isOpenrouter = data.npm === "@openrouter/ai-sdk-provider";
       const isAnthropic = data.npm === "@ai-sdk/anthropic";
+      const isKiro = data.npm === "kiro-acp-ai-provider";
       const hasApi = data.api !== undefined;
 
       return (
@@ -161,17 +216,20 @@ export const Provider = z
         isAnthropic ||
         // openai: api optional (always allowed)
         isOpenAI ||
+        // kiro: api optional (always allowed)
+        isKiro ||
         // all others: must NOT have api
         (!isOpenAI &&
           !isOpenAIcompatible &&
           !isOpenrouter &&
           !isAnthropic &&
+          !isKiro &&
           !hasApi)
       );
     },
     {
       message:
-        "'api' is required for openai-compatible and openrouter, optional for anthropic and openai, forbidden otherwise",
+        "'api' is required for openai-compatible and openrouter, optional for anthropic, openai, and kiro, forbidden otherwise",
       path: ["api"],
     },
   );

packages/web/script/build.ts

@@ -1,7 +1,6 @@
 #!/usr/bin/env bun
 
 import { Rendered, Providers } from "../src/render";
-import { normalizeLogoSvg } from "../src/logo.js";
 import fs from "fs/promises";
 import path from "path";
 import { $ } from "bun";
@@ -24,10 +23,7 @@ await fs.mkdir("./dist/logos", { recursive: true });
 const defaultLogoPath = "../../providers/logo.svg";
 const defaultLogo = Bun.file(defaultLogoPath);
 if (await defaultLogo.exists()) {
-  await Bun.write(
-    "./dist/logos/default.svg",
-    normalizeLogoSvg(await defaultLogo.text())
-  );
+  await Bun.write("./dist/logos/default.svg", defaultLogo);
 }
 
 // Then copy provider-specific logos
@@ -40,10 +36,7 @@ for (const entry of entries) {
     const logoFile = Bun.file(logoPath);
 
     if (await logoFile.exists()) {
-      await Bun.write(
-        `./dist/logos/${provider}.svg`,
-        normalizeLogoSvg(await logoFile.text())
-      );
+      await Bun.write(`./dist/logos/${provider}.svg`, logoFile);
     }
   }
 }

packages/web/src/index.css

@@ -319,12 +319,15 @@ tbody {
     gap: 0.375rem;
   }
 
-  .provider-logo {
+  .provider-cell span:first-child {
     flex: 0 0 auto;
+  }
+
+  .provider-cell svg {
     display: block;
     width: 1rem;
     height: 1rem;
-    object-fit: contain;
+    color: var(--color-text-secondary);
   }
 
   .model-id-cell {
@@ -413,17 +416,6 @@ tbody {
   .modality-icon:hover::after {
     opacity: 1;
   }
-
-  tr.loading-row td,
-  tr.error-row td {
-    padding: 1rem 0.75rem;
-    text-align: center;
-    font-family: inherit;
-    font-size: 0.875rem;
-    font-weight: 400;
-    text-transform: none;
-    color: var(--color-text-secondary);
-  }
 }
 
 dialog::backdrop {
@@ -554,4 +546,4 @@ dialog {
     }
   }
 
-}
+}

packages/web/src/index.ts

@@ -1,248 +1,7 @@
-interface ApiCost {
-  input?: number;
-  output?: number;
-  reasoning?: number;
-  cache_read?: number;
-  cache_write?: number;
-  input_audio?: number;
-  output_audio?: number;
-}
-
-interface ApiLimit {
-  context: number;
-  input?: number;
-  output: number;
-}
-
-interface ApiModel {
-  name: string;
-  family?: string;
-  status?: string;
-  tool_call: boolean;
-  reasoning: boolean;
-  modalities: {
-    input: string[];
-    output: string[];
-  };
-  cost?: ApiCost;
-  limit: ApiLimit;
-  structured_output?: boolean;
-  temperature: boolean;
-  open_weights: boolean;
-  knowledge?: string;
-  release_date: string;
-  last_updated: string;
-}
-
-interface ApiProvider {
-  name: string;
-  models: Record<string, ApiModel>;
-}
-
-type ApiResponse = Record<string, ApiProvider>;
-
-const COLUMN_COUNT = 25;
 const modal = document.getElementById("modal") as HTMLDialogElement;
 const modalClose = document.getElementById("close")!;
 const help = document.getElementById("help")!;
 const search = document.getElementById("search")! as HTMLInputElement;
-const tableBody = document.getElementById("table-body")! as HTMLTableSectionElement;
-
-const copyIcon = `
-  <svg
-    class="copy-icon"
-    xmlns="http://www.w3.org/2000/svg"
-    width="14"
-    height="14"
-    viewBox="0 0 24 24"
-    fill="none"
-    stroke="currentColor"
-    stroke-width="2"
-    stroke-linecap="round"
-    stroke-linejoin="round"
-  >
-    <rect width="14" height="14" x="8" y="8" rx="2" ry="2"></rect>
-    <path d="m4 16c-1.1 0-2-.9-2-2V4c0-1.1.9-2 2-2h10c1.1 0 2 .9 2 2"></path>
-  </svg>
-`;
-
-const checkIcon = `
-  <svg
-    class="check-icon"
-    xmlns="http://www.w3.org/2000/svg"
-    width="14"
-    height="14"
-    viewBox="0 0 24 24"
-    fill="none"
-    stroke="currentColor"
-    stroke-width="2"
-    stroke-linecap="round"
-    stroke-linejoin="round"
-    style="display: none;"
-  >
-    <polyline points="20,6 9,17 4,12"></polyline>
-  </svg>
-`;
-
-const modalityIcons: Record<string, { label: string; svg: string }> = {
-  text: {
-    label: "Text",
-    svg: `
-      <svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
-        <polyline points="4,7 4,4 20,4 20,7"></polyline>
-        <line x1="9" y1="20" x2="15" y2="20"></line>
-        <line x1="12" y1="4" x2="12" y2="20"></line>
-      </svg>
-    `,
-  },
-  image: {
-    label: "Image",
-    svg: `
-      <svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
-        <rect width="18" height="18" x="3" y="3" rx="2" ry="2"></rect>
-        <circle cx="9" cy="9" r="2"></circle>
-        <path d="m21 15-3.086-3.086a2 2 0 0 0-2.828 0L6 21"></path>
-      </svg>
-    `,
-  },
-  audio: {
-    label: "Audio",
-    svg: `
-      <svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
-        <polygon points="11 5 6 9 2 9 2 15 6 15 11 19 11 5"></polygon>
-        <path d="m19.07 4.93a10 10 0 0 1 0 14.14M15.54 8.46a5 5 0 0 1 0 7.07"></path>
-      </svg>
-    `,
-  },
-  video: {
-    label: "Video",
-    svg: `
-      <svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
-        <path d="m22 8-6 4 6 4V8Z"></path>
-        <rect width="14" height="12" x="2" y="6" rx="2" ry="2"></rect>
-      </svg>
-    `,
-  },
-  pdf: {
-    label: "PDF",
-    svg: `
-      <svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
-        <path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8z"></path>
-        <polyline points="14,2 14,8 20,8"></polyline>
-        <line x1="16" y1="13" x2="8" y2="13"></line>
-        <line x1="16" y1="17" x2="8" y2="17"></line>
-        <polyline points="10,9 9,9 8,9"></polyline>
-      </svg>
-    `,
-  },
-};
-
-function escapeHtml(value: string) {
-  return value
-    .replaceAll("&", "&amp;")
-    .replaceAll("<", "&lt;")
-    .replaceAll(">", "&gt;")
-    .replaceAll('"', "&quot;")
-    .replaceAll("'", "&#39;");
-}
-
-function renderCost(cost?: number) {
-  return cost === undefined ? "-" : `$${cost.toFixed(2)}`;
-}
-
-function renderModalityIcon(modality: string) {
-  const icon = modalityIcons[modality];
-  if (!icon) return "";
-
-  return `<span class="modality-icon" data-tooltip="${icon.label}">${icon.svg}</span>`;
-}
-
-function renderModalities(modalities: string[]) {
-  return `<div class="modalities">${modalities.map(renderModalityIcon).join("")}</div>`;
-}
-
-function renderProviderLogo(providerId: string) {
-  return `<img class="provider-logo" src="/logos/${encodeURIComponent(providerId)}.svg" alt="" width="16" height="16" loading="lazy" decoding="async" />`;
-}
-
-function renderRow(
-  providerId: string,
-  providerName: string,
-  modelId: string,
-  model: ApiModel
-) {
-  const safeProviderId = escapeHtml(providerId);
-  const safeProviderName = escapeHtml(providerName);
-  const safeModelId = escapeHtml(modelId);
-  const safeModelName = escapeHtml(model.name);
-  const safeFamily = escapeHtml(model.family ?? "-");
-  const safeKnowledge = escapeHtml(model.knowledge?.substring(0, 7) ?? "-");
-  const safeReleaseDate = escapeHtml(model.release_date);
-  const safeLastUpdated = escapeHtml(model.last_updated);
-
-  return `
-    <tr data-model-row="true">
-      <td>
-        <div class="provider-cell">
-          ${renderProviderLogo(providerId)}
-          <span>${safeProviderName}</span>
-        </div>
-      </td>
-      <td>${safeModelName}</td>
-      <td>${safeFamily}</td>
-      <td>${safeProviderId}</td>
-      <td>
-        <div class="model-id-cell">
-          <span class="model-id-text">${safeModelId}</span>
-          <button class="copy-button" type="button" data-model-id="${safeModelId}" aria-label="Copy model ID">
-            ${copyIcon}
-            ${checkIcon}
-          </button>
-        </div>
-      </td>
-      <td>${model.tool_call ? "Yes" : "No"}</td>
-      <td>${model.reasoning ? "Yes" : "No"}</td>
-      <td>${renderModalities(model.modalities.input)}</td>
-      <td>${renderModalities(model.modalities.output)}</td>
-      <td>${renderCost(model.cost?.input)}</td>
-      <td>${renderCost(model.cost?.output)}</td>
-      <td>${renderCost(model.cost?.reasoning)}</td>
-      <td>${renderCost(model.cost?.cache_read)}</td>
-      <td>${renderCost(model.cost?.cache_write)}</td>
-      <td>${renderCost(model.cost?.input_audio)}</td>
-      <td>${renderCost(model.cost?.output_audio)}</td>
-      <td>${model.limit.context.toLocaleString()}</td>
-      <td>${model.limit.input?.toLocaleString() ?? "-"}</td>
-      <td>${model.limit.output.toLocaleString()}</td>
-      <td>${model.structured_output === undefined ? "-" : model.structured_output ? "Yes" : "No"}</td>
-      <td>${model.temperature ? "Yes" : "No"}</td>
-      <td>${model.open_weights ? "Open" : "Closed"}</td>
-      <td>${safeKnowledge}</td>
-      <td>${safeReleaseDate}</td>
-      <td>${safeLastUpdated}</td>
-    </tr>
-  `;
-}
-
-function renderTableRows(providers: ApiResponse) {
-  return Object.entries(providers)
-    .sort(([, providerA], [, providerB]) =>
-      providerA.name.localeCompare(providerB.name)
-    )
-    .flatMap(([providerId, provider]) =>
-      Object.entries(provider.models)
-        .filter(([, model]) => model.status !== "alpha")
-        .sort(([, modelA], [, modelB]) => modelA.name.localeCompare(modelB.name))
-        .map(([modelId, model]) =>
-          renderRow(providerId, provider.name, modelId, model)
-        )
-    )
-    .join("");
-}
-
-function setStatusRow(message: string, className = "loading-row") {
-  tableBody.innerHTML = `<tr class="${className}"><td colspan="${COLUMN_COUNT}">${escapeHtml(message)}</td></tr>`;
-}
 
 /////////////////////////
 // URL State Management
@@ -251,10 +10,7 @@ function getQueryParams() {
   return new URLSearchParams(window.location.search);
 }
 
-function updateQueryParams(
-  updates: Record<string, string | null>,
-  historyMode: "push" | "replace" = "push"
-) {
+function updateQueryParams(updates: Record<string, string | null>) {
   const params = getQueryParams();
   for (const [key, value] of Object.entries(updates)) {
     if (value) {
@@ -263,18 +19,9 @@ function updateQueryParams(
       params.delete(key);
     }
   }
-
   const newPath = params.toString()
     ? `${window.location.pathname}?${params.toString()}`
     : window.location.pathname;
-
-  if (newPath === `${window.location.pathname}${window.location.search}`) return;
-
-  if (historyMode === "replace") {
-    window.history.replaceState({}, "", newPath);
-    return;
-  }
-
   window.history.pushState({}, "", newPath);
 }
 
@@ -318,71 +65,66 @@ modal.addEventListener("click", (e) => {
 ////////////////////
 // Handle Sorting
 ////////////////////
-let currentSort = { column: -1, direction: "asc" as "asc" | "desc" };
+let currentSort = { column: -1, direction: "asc" };
 
-function updateSortIndicators(column: number, direction: "asc" | "desc") {
-  const headers = document.querySelectorAll("th.sortable");
-  headers.forEach((header, i) => {
-    const indicator = header.querySelector(".sort-indicator")!;
-    indicator.textContent = i === column ? (direction === "asc" ? "↑" : "↓") : "";
-  });
-}
-
-function clearSortIndicators() {
-  updateSortIndicators(-1, "asc");
-}
-
-function sortTable(
-  column: number,
-  direction: "asc" | "desc",
-  syncUrl = true
-) {
+function sortTable(column: number, direction: "asc" | "desc") {
   const header = document.querySelectorAll("th.sortable")[column];
-  const columnType = header?.getAttribute("data-type");
+  const columnType = header.getAttribute("data-type");
   if (!columnType) return;
 
+  // update state
   currentSort = { column, direction };
-  if (syncUrl) {
-    updateQueryParams(
-      {
-        sort: getColumnNameForURL(header),
-        order: direction,
-      },
-      "push"
-    );
-  }
+  updateQueryParams({
+    sort: getColumnNameForURL(header),
+    order: direction,
+  });
 
+  // sort rows
+  const tbody = document.querySelector("table tbody")!;
   const rows = Array.from(
-    tableBody.querySelectorAll('tr[data-model-row="true"]')
+    tbody.querySelectorAll("tr")
   ) as HTMLTableRowElement[];
-
   rows.sort((a, b) => {
     const aValue = getCellValue(a.cells[column], columnType);
     const bValue = getCellValue(b.cells[column], columnType);
 
+    // Handle undefined values - always sort to bottom
     if (aValue === undefined && bValue === undefined) return 0;
     if (aValue === undefined) return 1;
     if (bValue === undefined) return -1;
 
-    const comparison =
-      columnType === "number" || columnType === "modalities"
-        ? (aValue as number) - (bValue as number)
-        : (aValue as string).localeCompare(bValue as string);
+    let comparison = 0;
+    if (columnType === "number" || columnType === "modalities") {
+      comparison = (aValue as number) - (bValue as number);
+    } else if (columnType === "boolean") {
+      comparison = (aValue as string).localeCompare(bValue as string);
+    } else {
+      comparison = (aValue as string).localeCompare(bValue as string);
+    }
 
     return direction === "asc" ? comparison : -comparison;
   });
+  rows.forEach((row) => tbody.appendChild(row));
 
-  rows.forEach((row) => tableBody.appendChild(row));
-  updateSortIndicators(column, direction);
+  // update sort indicators
+  const headers = document.querySelectorAll("th.sortable");
+  headers.forEach((header, i) => {
+    const indicator = header.querySelector(".sort-indicator")!;
+
+    if (i === column) {
+      indicator.textContent = direction === "asc" ? "↑" : "↓";
+    } else {
+      indicator.textContent = "";
+    }
+  });
 }
 
 function getCellValue(
   cell: HTMLTableCellElement,
   type: string
 ): string | number | undefined {
-  if (type === "modalities") {
+  if (type === "modalities")
     return cell.querySelectorAll(".modality-icon").length;
-  }
 
   const text = cell.textContent?.trim() || "";
   if (text === "-") return;
@@ -404,31 +146,22 @@ document.querySelectorAll("th.sortable").forEach((header) => {
 ///////////////////
 // Handle Search
 ///////////////////
-function filterTable(value: string, syncUrl = true) {
-  const lowerCaseValues = value
-    .toLowerCase()
-    .split(",")
-    .map((part) => part.trim())
-    .filter(Boolean);
-  const rows = tableBody.querySelectorAll(
-    'tr[data-model-row="true"]'
+function filterTable(value: string) {
+  const lowerCaseValues = value.toLowerCase().split(",").filter(str => str.trim() !== "");
+  const rows = document.querySelectorAll(
+    "table tbody tr"
   ) as NodeListOf<HTMLTableRowElement>;
 
   rows.forEach((row) => {
     const cellTexts = Array.from(row.cells).map((cell) =>
       cell.textContent!.toLowerCase()
     );
-    const isVisible =
-      lowerCaseValues.length === 0 ||
-      lowerCaseValues.some((lowerCaseValue) =>
-        cellTexts.some((text) => text.includes(lowerCaseValue))
-      );
+    const isVisible = lowerCaseValues.length === 0 ||
+     lowerCaseValues.some((lowerCaseValue) => cellTexts.some((text) => text.includes(lowerCaseValue)));
     row.style.display = isVisible ? "" : "none";
   });
 
-  if (syncUrl) {
-    updateQueryParams({ search: value || null }, "replace");
-  }
+  updateQueryParams({ search: value || null });
 }
 
 search.addEventListener("input", () => {
@@ -452,22 +185,22 @@ search.addEventListener("keydown", (e) => {
 ///////////////////////////////////
 // Handle Copy model ID function
 ///////////////////////////////////
-tableBody.addEventListener("click", async (event) => {
-  const target = event.target as HTMLElement;
-  const button = target.closest(".copy-button") as HTMLButtonElement | null;
-  const modelId = button?.dataset.modelId;
-  if (!button || !modelId) return;
-
+(window as any).copyModelId = async (
+  button: HTMLButtonElement,
+  modelId: string
+) => {
   try {
     if (navigator.clipboard) {
       await navigator.clipboard.writeText(modelId);
 
+      // Switch to check icon
       const copyIcon = button.querySelector(".copy-icon") as HTMLElement;
       const checkIcon = button.querySelector(".check-icon") as HTMLElement;
 
       copyIcon.style.display = "none";
       checkIcon.style.display = "block";
 
+      // Switch back after 1 second
       setTimeout(() => {
         copyIcon.style.display = "block";
         checkIcon.style.display = "none";
@@ -476,67 +209,32 @@ tableBody.addEventListener("click", async (event) => {
   } catch (err) {
     console.error("Failed to copy text: ", err);
   }
-});
+};
 
 ///////////////////////////////////
 // Initialize State from URL
 ///////////////////////////////////
-let tableLoaded = false;
-
 function initializeFromURL() {
-  if (!tableLoaded) return;
-
   const params = getQueryParams();
-  const searchQuery = params.get("search") ?? "";
-  search.value = searchQuery;
-  filterTable(searchQuery, false);
 
-  const columnName = params.get("sort");
-  if (!columnName) {
-    currentSort = { column: -1, direction: "asc" };
-    clearSortIndicators();
-    return;
-  }
+  (() => {
+    const searchQuery = params.get("search");
+    if (!searchQuery) return;
+    search.value = searchQuery;
+    filterTable(searchQuery);
+  })();
 
-  const columnIndex = getColumnIndexByUrlName(columnName);
-  if (columnIndex === -1) return;
+  (() => {
+    const columnName = params.get("sort");
+    if (!columnName) return;
 
-  const direction = (params.get("order") as "asc" | "desc") || "asc";
-  sortTable(columnIndex, direction, false);
+    const columnIndex = getColumnIndexByUrlName(columnName);
+    if (columnIndex === -1) return;
+
+    const direction = (params.get("order") as "asc" | "desc") || "asc";
+    sortTable(columnIndex, direction);
+  })();
 }
 
-async function loadTable() {
-  try {
-    const response = await fetch("/api.json");
-    if (!response.ok) {
-      throw new Error(`Failed to fetch models: ${response.status}`);
-    }
-
-    const providers = (await response.json()) as ApiResponse;
-    const rows = renderTableRows(providers);
-    tableBody.innerHTML = rows || "";
-
-    if (!rows) {
-      setStatusRow("No models found.");
-      return;
-    }
-
-    tableLoaded = true;
-    initializeFromURL();
-    window.addEventListener("popstate", initializeFromURL);
-  } catch (error) {
-    console.error("Failed to load model data:", error);
-    setStatusRow("Failed to load models.", "loading-row error-row");
-  }
-}
-
-function initializeApp() {
-  search.value = getQueryParams().get("search") ?? "";
-  void loadTable();
-}
-
-if (document.readyState === "loading") {
-  document.addEventListener("DOMContentLoaded", initializeApp);
-} else {
-  initializeApp();
-}
+document.addEventListener("DOMContentLoaded", initializeFromURL);
+window.addEventListener("popstate", initializeFromURL);

packages/web/src/logo.ts

@@ -1,16 +0,0 @@
-const LOGO_THEME_MARKER = "models-dev-logo-theme";
-const LOGO_THEME_STYLE = `<style id="${LOGO_THEME_MARKER}">:root{color:#666}@media (prefers-color-scheme: dark){:root{color:#AAA}}</style>`;
-
-export function normalizeLogoSvg(svgText: string) {
-  if (svgText.includes(LOGO_THEME_MARKER)) {
-    return svgText;
-  }
-
-  return svgText.replace(/<svg\b[^>]*>/i, (svgTag) => {
-    const themedTag = svgTag.includes("fill=")
-      ? svgTag
-      : svgTag.replace("<svg", '<svg fill="currentColor"');
-
-    return `${themedTag}${LOGO_THEME_STYLE}`;
-  });
-}

packages/web/src/render.tsx

@@ -1,15 +1,184 @@
 /** @jsx jsx */
 /** @jsxImportSource hono/jsx */
 
+import { generate } from "models.dev";
 import { Fragment } from "hono/jsx";
 import { renderToString } from "hono/jsx/dom/server";
-import { generate } from "models.dev";
+import { existsSync } from "fs";
 import path from "path";
 
 export const Providers = await generate(
   path.join(import.meta.dir, "..", "..", "..", "providers")
 );
 
+// Function to load SVG content
+const loadProviderSvg = async (providerId: string): Promise<string | null> => {
+  const providerLogoPath = path.join(
+    import.meta.dir,
+    "..",
+    "..",
+    "..",
+    "providers",
+    providerId,
+    "logo.svg"
+  );
+
+  const defaultLogoPath = path.join(
+    import.meta.dir,
+    "..",
+    "..",
+    "..",
+    "providers",
+    "logo.svg"
+  );
+
+  try {
+    // Try provider-specific logo first
+    if (existsSync(providerLogoPath)) {
+      const file = Bun.file(providerLogoPath);
+      return await file.text();
+    }
+    //
+    // Fall back to default logo
+    if (existsSync(defaultLogoPath)) {
+      const file = Bun.file(defaultLogoPath);
+      return await file.text();
+    }
+    return null;
+  } catch (error) {
+    console.warn(`Failed to load logo for provider ${providerId}:`, error);
+    return null;
+  }
+};
+
+// Create a cache of loaded SVGs at build time
+const providerLogos = new Map<string, string>();
+
+// Pre-load all provider logos
+for (const [providerId] of Object.entries(Providers)) {
+  const svgContent = await loadProviderSvg(providerId);
+  if (svgContent) {
+    providerLogos.set(providerId, svgContent);
+  }
+}
+
+function renderProviderLogo(providerId: string) {
+  const svgContent = providerLogos.get(providerId) || "";
+
+  return <span dangerouslySetInnerHTML={{ __html: svgContent }} />;
+}
+
+const getModalityIcon = (modality: string) => {
+  switch (modality) {
+    case "text":
+      return (
+        <span class="modality-icon" data-tooltip="Text">
+          <svg
+            xmlns="http://www.w3.org/2000/svg"
+            width="16"
+            height="16"
+            viewBox="0 0 24 24"
+            fill="none"
+            stroke="currentColor"
+            stroke-width="2"
+            stroke-linecap="round"
+            stroke-linejoin="round"
+          >
+            <polyline points="4,7 4,4 20,4 20,7"></polyline>
+            <line x1="9" y1="20" x2="15" y2="20"></line>
+            <line x1="12" y1="4" x2="12" y2="20"></line>
+          </svg>
+        </span>
+      );
+    case "image":
+      return (
+        <span class="modality-icon" data-tooltip="Image">
+          <svg
+            xmlns="http://www.w3.org/2000/svg"
+            width="16"
+            height="16"
+            viewBox="0 0 24 24"
+            fill="none"
+            stroke="currentColor"
+            stroke-width="2"
+            stroke-linecap="round"
+            stroke-linejoin="round"
+          >
+            <rect width="18" height="18" x="3" y="3" rx="2" ry="2"></rect>
+            <circle cx="9" cy="9" r="2"></circle>
+            <path d="m21 15-3.086-3.086a2 2 0 0 0-2.828 0L6 21"></path>
+          </svg>
+        </span>
+      );
+    case "audio":
+      return (
+        <span class="modality-icon" data-tooltip="Audio">
+          <svg
+            xmlns="http://www.w3.org/2000/svg"
+            width="16"
+            height="16"
+            viewBox="0 0 24 24"
+            fill="none"
+            stroke="currentColor"
+            stroke-width="2"
+            stroke-linecap="round"
+            stroke-linejoin="round"
+          >
+            <polygon points="11 5 6 9 2 9 2 15 6 15 11 19 11 5"></polygon>
+            <path d="m19.07 4.93a10 10 0 0 1 0 14.14M15.54 8.46a5 5 0 0 1 0 7.07"></path>
+          </svg>
+        </span>
+      );
+    case "video":
+      return (
+        <span class="modality-icon" data-tooltip="Video">
+          <svg
+            xmlns="http://www.w3.org/2000/svg"
+            width="16"
+            height="16"
+            viewBox="0 0 24 24"
+            fill="none"
+            stroke="currentColor"
+            stroke-width="2"
+            stroke-linecap="round"
+            stroke-linejoin="round"
+          >
+            <path d="m22 8-6 4 6 4V8Z"></path>
+            <rect width="14" height="12" x="2" y="6" rx="2" ry="2"></rect>
+          </svg>
+        </span>
+      );
+    case "pdf":
+      return (
+        <span class="modality-icon" data-tooltip="PDF">
+          <svg
+            xmlns="http://www.w3.org/2000/svg"
+            width="16"
+            height="16"
+            viewBox="0 0 24 24"
+            fill="none"
+            stroke="currentColor"
+            stroke-width="2"
+            stroke-linecap="round"
+            stroke-linejoin="round"
+          >
+            <path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8z"></path>
+            <polyline points="14,2 14,8 20,8"></polyline>
+            <line x1="16" y1="13" x2="8" y2="13"></line>
+            <line x1="16" y1="17" x2="8" y2="17"></line>
+            <polyline points="10,9 9,9 8,9"></polyline>
+          </svg>
+        </span>
+      );
+    default:
+      return null;
+  }
+};
+
+const renderCost = (cost?: number) => {
+  return cost === undefined ? "-" : `$${cost.toFixed(2)}`;
+};
+
 export const Rendered = renderToString(
   <Fragment>
     <header>
@@ -173,10 +342,118 @@ export const Rendered = renderToString(
           </th>
         </tr>
       </thead>
-      <tbody id="table-body">
-        <tr class="loading-row">
-          <td colspan={25}>Loading models...</td>
-        </tr>
+      <tbody>
+        {Object.entries(Providers)
+          .sort(([, providerA], [, providerB]) =>
+            providerA.name.localeCompare(providerB.name)
+          )
+          .flatMap(([providerId, provider]) =>
+            Object.entries(provider.models)
+              .filter(([, model]) => model.status !== "alpha")
+              .sort(([, modelA], [, modelB]) =>
+                modelA.name.localeCompare(modelB.name)
+              )
+              .map(([modelId, model]) => (
+                <tr key={`${providerId}-${modelId}`}>
+                  <td>
+                    <div class="provider-cell">
+                      {renderProviderLogo(providerId)}
+                      <span>{provider.name}</span>
+                    </div>
+                  </td>
+                  <td>{model.name}</td>
+                  <td>{model.family ?? "-"}</td>
+                  <td>{providerId}</td>
+                  <td>
+                    <div class="model-id-cell">
+                      <span class="model-id-text">{modelId}</span>
+                      <button
+                        class="copy-button"
+                        onclick={`copyModelId(this, '${modelId}')`}
+                      >
+                        <svg
+                          class="copy-icon"
+                          xmlns="http://www.w3.org/2000/svg"
+                          width="14"
+                          height="14"
+                          viewBox="0 0 24 24"
+                          fill="none"
+                          stroke="currentColor"
+                          stroke-width="2"
+                          stroke-linecap="round"
+                          stroke-linejoin="round"
+                        >
+                          <rect
+                            width="14"
+                            height="14"
+                            x="8"
+                            y="8"
+                            rx="2"
+                            ry="2"
+                          />
+                          <path d="m4 16c-1.1 0-2-.9-2-2V4c0-1.1.9-2 2-2h10c1.1 0 2 .9 2 2" />
+                        </svg>
+                        <svg
+                          class="check-icon"
+                          xmlns="http://www.w3.org/2000/svg"
+                          width="14"
+                          height="14"
+                          viewBox="0 0 24 24"
+                          fill="none"
+                          stroke="currentColor"
+                          stroke-width="2"
+                          stroke-linecap="round"
+                          stroke-linejoin="round"
+                          style="display: none;"
+                        >
+                          <polyline points="20,6 9,17 4,12" />
+                        </svg>
+                      </button>
+                    </div>
+                  </td>
+                  <td>{model.tool_call ? "Yes" : "No"}</td>
+                  <td>{model.reasoning ? "Yes" : "No"}</td>
+                  <td>
+                    <div class="modalities">
+                      {model.modalities.input.map((modality) =>
+                        getModalityIcon(modality)
+                      )}
+                    </div>
+                  </td>
+                  <td>
+                    <div class="modalities">
+                      {model.modalities.output.map((modality) =>
+                        getModalityIcon(modality)
+                      )}
+                    </div>
+                  </td>
+                  <td>{renderCost(model.cost?.input)}</td>
+                  <td>{renderCost(model.cost?.output)}</td>
+                  <td>{renderCost(model.cost?.reasoning)}</td>
+                  <td>{renderCost(model.cost?.cache_read)}</td>
+                  <td>{renderCost(model.cost?.cache_write)}</td>
+                  <td>{renderCost(model.cost?.input_audio)}</td>
+                  <td>{renderCost(model.cost?.output_audio)}</td>
+                  <td>{model.limit.context.toLocaleString()}</td>
+                  <td>{model.limit.input?.toLocaleString() ?? "-"}</td>
+                  <td>{model.limit.output.toLocaleString()}</td>
+                  <td>
+                    {model.structured_output === undefined
+                      ? "-"
+                      : model.structured_output
+                      ? "Yes"
+                      : "No"}
+                  </td>
+                  <td>{model.temperature ? "Yes" : "No"}</td>
+                  <td>{model.open_weights ? "Open" : "Closed"}</td>
+                  <td>
+                    {model.knowledge ? model.knowledge.substring(0, 7) : "-"}
+                  </td>
+                  <td>{model.release_date}</td>
+                  <td>{model.last_updated}</td>
+                </tr>
+              ))
+          )}
       </tbody>
     </table>
     <dialog id="modal">

packages/web/src/server.ts

@@ -1,19 +1,11 @@
 import Index from "../index.html";
-import { normalizeLogoSvg } from "./logo.js";
-import { Providers, Rendered } from "./render";
+import { Rendered } from "./render";
 import path from "path";
 
 Bun.serve({
   port: 16_000,
   routes: {
     "/": Index,
-    "/api.json": () => {
-      return Response.json(Providers, {
-        headers: {
-          "Cache-Control": "public, max-age=3600",
-        },
-      });
-    },
     "/assets/*": (req) => {
       const file = Bun.file(
         path.join(import.meta.dir, new URL(req.url).pathname)
@@ -46,7 +38,7 @@ Bun.serve({
         file = Bun.file(defaultLogoPath);
       }
 
-      return new Response(normalizeLogoSvg(await file.text()), {
+      return new Response(file, {
         headers: {
           "Content-Type": "image/svg+xml",
           "Cache-Control": "public, max-age=3600",

providers/302ai/models/claude-3-5-haiku-20241022.toml

@@ -1,4 +1,5 @@
 name = "claude-3-5-haiku-20241022"
+family = "claude-haiku"
 release_date = "2024-10-22"
 last_updated = "2024-10-22"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = false
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2024-07"
+knowledge = "2024-07-31"
 
 [cost]
 input = 0.800

providers/302ai/models/claude-3-5-haiku-latest.toml

@@ -1,4 +1,5 @@
 name = "claude-3-5-haiku-latest"
+family = "claude-haiku"
 release_date = "2024-10-22"
 last_updated = "2024-10-22"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = false
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2024-07"
+knowledge = "2024-07-31"
 
 [cost]
 input = 0.800

providers/302ai/models/claude-haiku-4-5-20251001.toml

@@ -1,4 +1,5 @@
 name = "claude-haiku-4-5-20251001"
+family = "claude-haiku"
 release_date = "2025-10-16"
 last_updated = "2025-10-16"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-03"
+knowledge = "2025-02-28"
 
 [cost]
 input = 1.000

providers/302ai/models/claude-haiku-4-5.toml

@@ -1,4 +1,5 @@
 name = "claude-haiku-4-5"
+family = "claude-haiku"
 release_date = "2025-10-16"
 last_updated = "2025-10-16"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-02"
+knowledge = "2025-02-28"
 
 [cost]
 input = 1.000

providers/302ai/models/claude-opus-4-1-20250805.toml

@@ -1,12 +1,13 @@
 name = "claude-opus-4-1-20250805"
+family = "claude-opus"
 release_date = "2025-08-05"
 last_updated = "2025-08-05"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-03"
+knowledge = "2025-03-31"
 
 [cost]
 input = 15.000
@@ -17,5 +18,5 @@ context = 200_000
 output = 32_000
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "pdf"]
 output = ["text"]

providers/302ai/models/claude-opus-4-20250514.toml

@@ -1,12 +1,13 @@
 name = "claude-opus-4-20250514"
+family = "claude-opus"
 release_date = "2025-05-22"
 last_updated = "2025-05-22"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-03"
+knowledge = "2025-03-31"
 
 [cost]
 input = 15.000

providers/302ai/models/claude-opus-4-5-20251101.toml

@@ -1,4 +1,5 @@
 name = "claude-opus-4-5-20251101"
+family = "claude-opus"
 release_date = "2025-11-25"
 last_updated = "2025-11-25"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-03"
+knowledge = "2025-03-31"
 
 [cost]
 input = 5.000

providers/302ai/models/claude-opus-4-5.toml

@@ -1,4 +1,5 @@
 name = "claude-opus-4-5"
+family = "claude-opus"
 release_date = "2025-11-25"
 last_updated = "2025-11-25"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-03"
+knowledge = "2025-03-31"
 
 [cost]
 input = 5.000

providers/302ai/models/claude-opus-4-6.toml

@@ -1,4 +1,5 @@
 name = "claude-opus-4-6"
+family = "claude-opus"
 release_date = "2026-02-06"
 last_updated = "2026-03-13"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-05"
+knowledge = "2025-05-31"
 
 [cost]
 input = 5.000

providers/302ai/models/claude-opus-4-7.toml

@@ -1,4 +1,5 @@
 name = "claude-opus-4-7"
+family = "claude-opus"
 release_date = "2026-04-17"
 last_updated = "2026-04-17"
 attachment = true
@@ -6,7 +7,7 @@ reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-05"
+knowledge = "2026-01-31"
 
 [cost]
 input = 5.000
@@ -14,14 +15,15 @@ output = 25.000
 cache_read = 0.500
 cache_write = 6.250
 
-[cost.context_over_200k]
+[[cost.tiers]]
+tier = { size = 200_000 }
 input = 10.000
 output = 37.500
 cache_read = 1.000
 cache_write = 12.500
 
 [limit]
-context = 200_000
+context = 1_000_000
 output = 128_000
 
 [modalities]

providers/302ai/models/claude-sonnet-4-20250514.toml

@@ -1,12 +1,13 @@
 name = "claude-sonnet-4-20250514"
+family = "claude-sonnet"
 release_date = "2025-05-22"
 last_updated = "2025-05-22"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-03"
+knowledge = "2025-03-31"
 
 [cost]
 input = 3.000

providers/302ai/models/claude-sonnet-4-5-20250929.toml

@@ -1,12 +1,13 @@
 name = "claude-sonnet-4-5-20250929"
+family = "claude-sonnet"
 release_date = "2025-09-30"
 last_updated = "2025-09-30"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-03"
+knowledge = "2025-07-31"
 
 [cost]
 input = 3.000

providers/302ai/models/claude-sonnet-4-5.toml

@@ -1,12 +1,13 @@
 name = "claude-sonnet-4-5"
+family = "claude-sonnet"
 release_date = "2025-09-30"
 last_updated = "2025-09-30"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-07"
+knowledge = "2025-07-31"
 
 [cost]
 input = 3.000

providers/302ai/models/claude-sonnet-4-6.toml

@@ -1,12 +1,13 @@
 name = "claude-sonnet-4-6"
+family = "claude-sonnet"
 release_date = "2026-02-18"
 last_updated = "2026-03-13"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 open_weights = false
-knowledge = "2025-08"
+knowledge = "2025-08-31"
 
 [cost]
 input = 3.000

providers/302ai/models/glm-4.5-air.toml

@@ -1,11 +1,12 @@
 name = "glm-4.5-air"
+family = "glm-air"
 release_date = "2025-07-29"
 last_updated = "2025-07-29"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
+open_weights = true
 knowledge = "2025-04"
 
 [cost]
@@ -13,7 +14,7 @@ input = 0.1143
 output = 0.286
 
 [limit]
-context = 128_000
+context = 131_072
 output = 98_304
 
 [modalities]

providers/302ai/models/glm-4.5-airx.toml

@@ -1,4 +1,5 @@
 name = "glm-4.5-airx"
+family = "glm"
 release_date = "2025-07-29"
 last_updated = "2025-07-29"
 attachment = false

providers/302ai/models/glm-4.5-x.toml

@@ -1,4 +1,5 @@
 name = "glm-4.5-x"
+family = "glm"
 release_date = "2025-07-29"
 last_updated = "2025-07-29"
 attachment = false

providers/302ai/models/glm-4.5.toml

@@ -1,19 +1,20 @@
 name = "GLM-4.5"
+family = "glm"
 release_date = "2025-07-29"
 last_updated = "2025-07-29"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
-knowledge = "2024-10"
+open_weights = true
+knowledge = "2025-04"
 
 [cost]
 input = 0.286
 output = 1.142
 
 [limit]
-context = 128_000
+context = 131_072
 output = 98_304
 
 [modalities]

providers/302ai/models/glm-4.5v.toml

@@ -1,12 +1,13 @@
 name = "GLM-4.5V"
+family = "glm"
 release_date = "2025-08-12"
 last_updated = "2025-08-12"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
-knowledge = "2024-10"
+open_weights = true
+knowledge = "2025-04"
 
 [cost]
 input = 0.290
@@ -17,5 +18,5 @@ context = 64_000
 output = 16_384
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video"]
 output = ["text"]

providers/302ai/models/glm-4.6.toml

@@ -1,19 +1,20 @@
 name = "glm-4.6"
+family = "glm"
 release_date = "2025-09-30"
 last_updated = "2025-09-30"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
-knowledge = "2025-03"
+open_weights = true
+knowledge = "2025-04"
 
 [cost]
 input = 0.286
 output = 1.142
 
 [limit]
-context = 200_000
+context = 204_800
 output = 131_072
 
 [modalities]

providers/302ai/models/glm-4.6v.toml

@@ -1,12 +1,13 @@
 name = "GLM-4.6V"
+family = "glm"
 release_date = "2025-12-08"
 last_updated = "2025-12-08"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
-knowledge = "2025-03"
+open_weights = true
+knowledge = "2025-04"
 
 [cost]
 input = 0.145
@@ -17,5 +18,5 @@ context = 128_000
 output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "video"]
 output = ["text"]

providers/302ai/models/glm-4.7-flashx.toml

@@ -1,11 +1,12 @@
 name = "glm-4.7-flashx"
+family = "glm-flash"
 release_date = "2026-01-20"
 last_updated = "2026-01-20"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
+open_weights = true
 knowledge = "2025-04"
 
 [cost]

providers/302ai/models/glm-4.7.toml

@@ -1,19 +1,23 @@
 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
-open_weights = false
-knowledge = "2025-06"
+open_weights = true
+knowledge = "2025-04"
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.286
 output = 1.142
 
 [limit]
-context = 200_000
+context = 204_800
 output = 131_072
 
 [modalities]

providers/302ai/models/glm-5-turbo.toml

@@ -1,12 +1,17 @@
 name = "glm-5-turbo"
+family = "glm"
 release_date = "2026-03-16"
 last_updated = "2026-03-16"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
 input = 0.720
 output = 3.200

providers/302ai/models/glm-5.1.toml

@@ -1,12 +1,17 @@
 name = "glm-5.1"
+family = "glm"
 release_date = "2026-04-10"
 last_updated = "2026-04-10"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
 input = 0.860
 output = 3.500

providers/302ai/models/glm-5.toml

@@ -1,18 +1,22 @@
 name = "glm-5"
+family = "glm"
 release_date = "2026-02-12"
 last_updated = "2026-02-12"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
 input = 0.600
 output = 2.600
 
 [limit]
-context = 200_000
+context = 204_800
 output = 131_072
 
 [modalities]

providers/302ai/models/glm-5v-turbo.toml

@@ -1,4 +1,5 @@
-name = "glm-5v-turbo"
+name = "GLM-5V-Turbo"
+family = "glm"
 release_date = "2026-04-02"
 last_updated = "2026-04-02"
 attachment = true
@@ -7,6 +8,9 @@ temperature = true
 tool_call = true
 open_weights = false
 
+[interleaved]
+field = "reasoning_content"
+
 [cost]
 input = 0.720
 output = 3.200

providers/302ai/models/glm-for-coding.toml

@@ -1,4 +1,5 @@
 name = "glm-for-coding"
+family = "glm"
 release_date = "2025-09-30"
 last_updated = "2025-09-30"
 attachment = false

providers/302ai/models/gpt-4.1-mini.toml

@@ -6,6 +6,7 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 knowledge = "2024-04"
 
@@ -14,9 +15,9 @@ input = 0.400
 output = 1.600
 
 [limit]
-context = 1_000_000
+context = 1_047_576
 output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "pdf"]
 output = ["text"]

providers/302ai/models/gpt-4.1-nano.toml

@@ -6,6 +6,7 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 knowledge = "2024-04"
 
@@ -14,7 +15,7 @@ input = 0.100
 output = 0.400
 
 [limit]
-context = 1_000_000
+context = 1_047_576
 output = 32_768
 
 [modalities]

providers/302ai/models/gpt-4.1.toml

@@ -6,6 +6,7 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 knowledge = "2024-04"
 
@@ -14,9 +15,9 @@ input = 2.000
 output = 8.000
 
 [limit]
-context = 1_000_000
+context = 1_047_576
 output = 32_768
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "pdf"]
 output = ["text"]

providers/302ai/models/gpt-4o.toml

@@ -6,6 +6,7 @@ attachment = true
 reasoning = false
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 knowledge = "2023-09"
 
@@ -18,5 +19,5 @@ context = 128_000
 output = 16_384
 
 [modalities]
-input = ["text", "image"]
+input = ["text", "image", "pdf"]
 output = ["text"]

providers/302ai/models/gpt-5-mini.toml

@@ -1,12 +1,14 @@
 name = "gpt-5-mini"
+family = "gpt-mini"
 release_date = "2025-08-08"
 last_updated = "2025-08-08"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2024-10"
+knowledge = "2024-05-30"
 
 [cost]
 input = 0.250
@@ -14,6 +16,7 @@ output = 2.000
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/302ai/models/gpt-5-pro.toml

@@ -1,12 +1,14 @@
 name = "gpt-5-pro"
+family = "gpt-pro"
 release_date = "2025-10-08"
 last_updated = "2025-10-08"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2024-10"
+knowledge = "2024-09-30"
 
 [cost]
 input = 15.000
@@ -14,6 +16,7 @@ output = 120.000
 
 [limit]
 context = 400_000
+input = 272_000
 output = 272_000
 
 [modalities]

providers/302ai/models/gpt-5.1-chat-latest.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.1-chat-latest"
+family = "gpt-codex"
 release_date = "2025-11-14"
 last_updated = "2025-11-14"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2024-10"
+knowledge = "2024-09-30"
 
 [cost]
 input = 1.250

providers/302ai/models/gpt-5.1.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.1"
+family = "gpt"
 release_date = "2025-11-14"
 last_updated = "2025-11-14"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2024-10"
+knowledge = "2024-09-30"
 
 [cost]
 input = 1.250
@@ -14,6 +16,7 @@ output = 10.000
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/302ai/models/gpt-5.2-chat-latest.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.2-chat-latest"
+family = "gpt-codex"
 release_date = "2025-12-12"
 last_updated = "2025-12-12"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2024-10"
+knowledge = "2025-08-31"
 
 [cost]
 input = 1.750

providers/302ai/models/gpt-5.2.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.2"
+family = "gpt"
 release_date = "2025-12-12"
 last_updated = "2025-12-12"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2024-10"
+knowledge = "2025-08-31"
 
 [cost]
 input = 1.750
@@ -14,6 +16,7 @@ output = 14.000
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/302ai/models/gpt-5.4-mini-2026-03-17.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.4-mini-2026-03-17"
+family = "gpt-mini"
 release_date = "2026-03-19"
 last_updated = "2026-03-19"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2025-08"
+knowledge = "2025-08-31"
 
 [cost]
 input = 0.750
@@ -14,6 +16,7 @@ output = 4.500
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/302ai/models/gpt-5.4-mini.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.4-mini"
+family = "gpt-mini"
 release_date = "2026-03-19"
 last_updated = "2026-03-19"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2025-08"
+knowledge = "2025-08-31"
 
 [cost]
 input = 0.750
@@ -14,6 +16,7 @@ output = 4.500
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/302ai/models/gpt-5.4-nano-2026-03-17.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.4-nano-2026-03-17"
+family = "gpt-nano"
 release_date = "2026-03-19"
 last_updated = "2026-03-19"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2025-08"
+knowledge = "2025-08-31"
 
 [cost]
 input = 0.200
@@ -14,6 +16,7 @@ output = 1.250
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/302ai/models/gpt-5.4-nano.toml

@@ -1,12 +1,14 @@
 name = "gpt-5.4-nano"
+family = "gpt-nano"
 release_date = "2026-03-19"
 last_updated = "2026-03-19"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2025-08"
+knowledge = "2025-08-31"
 
 [cost]
 input = 0.200
@@ -14,6 +16,7 @@ output = 1.250
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/302ai/models/gpt-5.4-pro.toml

@@ -0,0 +1,31 @@
+name = "gpt-5.4-pro"
+family = "gpt-pro"
+release_date = "2026-03-05"
+last_updated = "2026-03-05"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = false
+open_weights = false
+knowledge = "2025-08-31"
+
+[cost]
+input = 30.000
+output = 180.000
+cache_read = 0
+cache_write = 0
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 60.000
+output = 270.000
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/302ai/models/gpt-5.4.toml

@@ -0,0 +1,31 @@
+name = "gpt-5.4"
+family = "gpt"
+release_date = "2026-03-05"
+last_updated = "2026-03-05"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+open_weights = false
+knowledge = "2025-08-31"
+
+[cost]
+input = 2.500
+output = 15.000
+cache_read = 0.250
+cache_write = 0
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 5.000
+output = 22.500
+
+[limit]
+context = 1_050_000
+input = 922_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

providers/302ai/models/gpt-5.toml

@@ -1,12 +1,14 @@
 name = "gpt-5"
+family = "gpt"
 release_date = "2025-08-08"
 last_updated = "2025-08-08"
 attachment = true
-reasoning = false
-temperature = true
+reasoning = true
+temperature = false
 tool_call = true
+structured_output = true
 open_weights = false
-knowledge = "2024-10"
+knowledge = "2024-09-30"
 
 [cost]
 input = 1.250
@@ -14,6 +16,7 @@ output = 10.000
 
 [limit]
 context = 400_000
+input = 272_000
 output = 128_000
 
 [modalities]

providers/abliteration-ai/logo.svg

@@ -0,0 +1,6 @@
+<svg viewBox="0 0 220 32" xmlns="http://www.w3.org/2000/svg">
+  <text x="0" y="24" fill="currentColor">
+    <tspan font-size="24">Abliteration</tspan>
+    <tspan dx="8" dy="2" font-size="12" fill-opacity="0.6">.ai</tspan>
+  </text>
+</svg>

providers/abliteration-ai/models/abliterated-model.toml

@@ -0,0 +1,22 @@
+name = "Abliterated Model"
+release_date = "2026-01-06"
+last_updated = "2026-01-06"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = false
+temperature = true
+open_weights = true
+
+[cost]
+input = 3.00
+output = 3.00
+
+[limit]
+context = 150_000
+input = 150_000
+output = 8_192
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/abliteration-ai/provider.toml

@@ -0,0 +1,5 @@
+name = "abliteration.ai"
+env = ["ABLIT_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://api.abliteration.ai/v1"
+doc = "https://docs.abliteration.ai/models"

providers/aihubmix/models/claude-opus-4-5.toml

@@ -1,23 +0,0 @@
-name = "Claude Opus 4.5"
-release_date = "2025-11-25"
-last_updated = "2025-11-25"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-knowledge = "2025-03"
-open_weights = false
-
-[cost]
-input = 5.00
-output = 25.00
-cache_read = 0.50
-cache_write = 6.25
-
-[limit]
-context = 200_000
-output = 32_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

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

@@ -1,7 +1,7 @@
-name = "Claude Opus 4.6 Think"
+name = "Claude Opus 4.6"
 family = "claude-opus"
 release_date = "2026-02-05"
-last_updated = "2026-02-05"
+last_updated = "2026-03-13"
 attachment = true
 reasoning = true
 temperature = true
@@ -10,20 +10,14 @@ knowledge = "2025-05-31"
 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
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
 context = 200_000
-output = 128_000
+output = 32_000
 
 [modalities]
 input = ["text", "image", "pdf"]

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

@@ -1,7 +1,7 @@
 name = "Claude Opus 4.6"
 family = "claude-opus"
 release_date = "2026-02-05"
-last_updated = "2026-02-05"
+last_updated = "2026-03-13"
 attachment = true
 reasoning = true
 temperature = true
@@ -10,19 +10,13 @@ knowledge = "2025-05-31"
 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
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
-context = 200_000
+context = 1_000_000
 output = 128_000
 
 [modalities]

providers/aihubmix/models/claude-opus-4-7-think.toml

@@ -0,0 +1,25 @@
+name = "Claude Opus 4.7 Thinking"
+family = "claude-opus"
+release_date = "2026-04-16"
+last_updated = "2026-04-16"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2026-01-31"
+open_weights = false
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
+
+[limit]
+context = 200_000
+output = 32_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/aihubmix/models/claude-opus-4-7.toml

@@ -0,0 +1,24 @@
+name = "Claude Opus 4.7"
+family = "claude-opus"
+release_date = "2026-04-16"
+last_updated = "2026-04-16"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+knowledge = "2026-01-31"
+open_weights = false
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
+
+[limit]
+context = 1_000_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]

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

@@ -15,7 +15,8 @@ output = 15.00
 cache_read = 0.30
 cache_write = 3.75
 
-[cost.context_over_200k]
+[[cost.tiers]]
+tier = { size = 200_000 }
 input = 6.00
 output = 22.50
 cache_read = 0.60

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

@@ -15,7 +15,8 @@ output = 15.00
 cache_read = 0.30
 cache_write = 3.75
 
-[cost.context_over_200k]
+[[cost.tiers]]
+tier = { size = 200_000 }
 input = 6.00
 output = 22.50
 cache_read = 0.60

providers/aihubmix/models/coding-glm-4.7-free.toml

@@ -1,28 +0,0 @@
-name = "Coding GLM 4.7 Free"
-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_details"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-cache_write = 0
-
-[limit]
-context = 204800
-output = 131072
-
-[modalities]
-input = ["text"]
-output = ["text"]
-

providers/aihubmix/models/coding-glm-4.7.toml

@@ -1,27 +0,0 @@
-name = "Coding-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_details"
-
-[cost]
-input = 0.06
-output = 0.22
-cache_read = 0.01
-
-[limit]
-context = 204800
-output = 131072
-
-[modalities]
-input = ["text"]
-output = ["text"]
-

providers/aihubmix/models/coding-glm-5.1-free.toml

@@ -1,25 +1,26 @@
-name = "GLM 4.5 TEE"
+name = "Coding GLM 5.1 (free)"
 family = "glm"
-release_date = "2025-12-29"
-last_updated = "2026-01-10"
+release_date = "2026-04-11"
+last_updated = "2026-04-11"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-open_weights = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
 
 [cost]
-input = 0.35
-output = 1.55
+input = 0
+output = 0
+cache_read = 0
 
 [limit]
-context = 131_072
-output = 65_536
+context = 204_800
+output = 128_000
 
 [modalities]
 input = ["text"]
 output = ["text"]
-
-[interleaved]
-field = "reasoning_content"

providers/aihubmix/models/coding-minimax-m2.7-free.toml

@@ -1,7 +1,7 @@
-name = "MiniMax-M2.5"
+name = "Coding-MiniMax-M2.7-Free"
 family = "minimax"
-release_date = "2026-02-12"
-last_updated = "2026-02-12"
+release_date = "2026-03-18"
+last_updated = "2026-03-18"
 attachment = false
 reasoning = true
 temperature = true
@@ -9,12 +9,12 @@ tool_call = true
 open_weights = true
 
 [cost]
-input = 0.29
-output = 1.15
+input = 0
+output = 0
 
 [limit]
 context = 204_800
-output = 131_072
+output = 13_100
 
 [modalities]
 input = ["text"]

providers/aihubmix/models/coding-minimax-m2.7-highspeed.toml

@@ -1,25 +1,22 @@
-name = "MiniMax M2.1"
+name = "Coding MiniMax M2.7 Highspeed"
 family = "minimax"
+release_date = "2026-03-18"
+last_updated = "2026-03-18"
 attachment = false
 reasoning = true
-tool_call = true
 temperature = true
-release_date = "2025-12-23"
-last_updated = "2025-12-23"
+tool_call = true
+structured_output = true
 open_weights = true
 
-[interleaved]
-field = "reasoning_details"
-
 [cost]
-input = 0.29
-output = 1.15
+input = 0.2
+output = 0.2
 
 [limit]
 context = 204_800
-output = 131_072
+output = 13_100
 
 [modalities]
 input = ["text"]
 output = ["text"]
-

providers/aihubmix/models/coding-minimax-m2.7.toml

@@ -1,20 +1,21 @@
-name = "MiniMax-M2.1"
+name = "Coding MiniMax M2.7"
 family = "minimax"
-release_date = "2025-12-23"
-last_updated = "2025-12-23"
+release_date = "2026-03-18"
+last_updated = "2026-03-18"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = true
 
 [cost]
-input = 0.0
-output = 0.0
+input = 0.2
+output = 0.2
 
 [limit]
 context = 204_800
-output = 131_072
+output = 13_100
 
 [modalities]
 input = ["text"]

providers/aihubmix/models/deepseek-v3.2-fast.toml

@@ -1,21 +0,0 @@
-name = "DeepSeek-V3.2-Fast"
-family = "deepseek"
-release_date = "2025-12-01"
-last_updated = "2025-12-01"
-attachment = false
-reasoning = false
-knowledge = "2024-07"
-tool_call = false
-open_weights = true
-
-[cost]
-input = 1.10
-output = 3.29
-
-[limit]
-context = 128_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

providers/aihubmix/models/deepseek-v4-flash-think.toml

@@ -0,0 +1,27 @@
+name = "DeepSeek V4 Flash Think"
+family = "deepseek"
+release_date = "2026-04-24"
+last_updated = "2026-04-24"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-05"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.154
+output = 0.308
+cache_read = 0.0308
+
+[limit]
+context = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/deepseek-v4-flash.toml

@@ -0,0 +1,27 @@
+name = "DeepSeek V4 Flash"
+family = "deepseek-flash"
+release_date = "2026-04-24"
+last_updated = "2026-04-24"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-05"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.154
+output = 0.308
+cache_read = 0.0308
+
+[limit]
+context = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/deepseek-v4-pro.toml

@@ -0,0 +1,27 @@
+name = "DeepSeek V4 Pro"
+family = "deepseek-thinking"
+release_date = "2026-04-24"
+last_updated = "2026-04-24"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-05"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.478
+output = 0.956
+cache_read = 0.004302
+
+[limit]
+context = 1_000_000
+output = 384_000
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/aihubmix/models/gemini-2.5-flash.toml

@@ -1,23 +1,24 @@
 name = "Gemini 2.5 Flash"
 family = "gemini-flash"
-release_date = "2025-09-15"
-last_updated = "2025-09-15"
+release_date = "2025-03-20"
+last_updated = "2025-06-05"
 attachment = true
-reasoning = false
+reasoning = true
 temperature = true
-knowledge = "2025-04"
 tool_call = true
+structured_output = true
+knowledge = "2025-01"
 open_weights = false
 
 [cost]
-input = 0.075
-output = 0.30
-cache_read = 0.02
+input = 0.3
+output = 2.499
+cache_read = 0.03
 
 [limit]
-context = 1_000_000
-output = 65_000
+context = 1_048_576
+output = 65_536
 
 [modalities]
-input = ["text", "image", "audio", "video"]
+input = ["text", "image", "audio", "video", "pdf"]
 output = ["text"]

providers/aihubmix/models/gemini-2.5-pro.toml

@@ -1,23 +1,24 @@
 name = "Gemini 2.5 Pro"
 family = "gemini-pro"
-release_date = "2025-09-15"
-last_updated = "2025-09-15"
+release_date = "2025-03-20"
+last_updated = "2025-06-05"
 attachment = true
 reasoning = true
 temperature = true
-knowledge = "2025-04"
 tool_call = true
+structured_output = true
+knowledge = "2025-01"
 open_weights = false
 
 [cost]
 input = 1.25
-output = 5.00
-cache_read = 0.31
+output = 10
+cache_read = 0.125
 
 [limit]
-context = 2_000_000
-output = 65_000
+context = 1_048_576
+output = 65_536
 
 [modalities]
-input = ["text", "image", "audio", "video"]
+input = ["text", "image", "audio", "video", "pdf"]
 output = ["text"]

providers/aihubmix/models/gemini-3-flash-preview.toml

@@ -0,0 +1,24 @@
+name = "Gemini 3 Flash Preview"
+family = "gemini-flash"
+release_date = "2025-12-17"
+last_updated = "2025-12-17"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-01"
+open_weights = false
+
+[cost]
+input = 0.5
+output = 3
+cache_read = 0.05
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]

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

@@ -1,23 +0,0 @@
-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/gemini-3-pro-preview.toml

@@ -1,23 +0,0 @@
-name = "Gemini 3 Pro Preview"
-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/gemini-3.1-flash-lite.toml

@@ -0,0 +1,24 @@
+name = "Gemini 3.1 Flash Lite"
+family = "gemini-flash-lite"
+release_date = "2026-03-03"
+last_updated = "2026-03-03"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+knowledge = "2025-01"
+open_weights = false
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.25
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]

providers/aihubmix/models/gemini-3.1-pro-preview-customtools.toml

@@ -0,0 +1,30 @@
+name = "Gemini 3.1 Pro Preview Custom Tools"
+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
+output = 12
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { size = 200_000 }
+input = 4
+output = 18
+cache_read = 0.4
+
+[limit]
+context = 1_048_576
+output = 65_536
+
+[modalities]
+input = ["text", "image", "audio", "video", "pdf"]
+output = ["text"]

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

@@ -1,7 +1,7 @@
-name = "Gemini 3 Pro Preview"
+name = "Gemini 3.1 Pro Preview"
 family = "gemini-pro"
-release_date = "2025-11-18"
-last_updated = "2025-11-18"
+release_date = "2026-02-19"
+last_updated = "2026-02-19"
 attachment = true
 reasoning = true
 temperature = true
@@ -16,9 +16,9 @@ output = 12
 cache_read = 0.2
 
 [limit]
-context = 1_000_000
-output = 64_000
+context = 1_048_576
+output = 65_536
 
 [modalities]
-input = ["text", "image", "video", "audio", "pdf"]
+input = ["text", "image", "audio", "video", "pdf"]
 output = ["text"]

providers/aihubmix/models/glm-4.6v.toml

@@ -1,22 +0,0 @@
-name = "GLM-4.6V"
-family = "glm"
-release_date = "2025-12-08"
-last_updated = "2025-12-08"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-knowledge = "2025-04"
-open_weights = true
-
-[cost]
-input = 0.14
-output = 0.41
-
-[limit]
-context = 128000
-output = 32768
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]

providers/aihubmix/models/glm-4.7.toml

@@ -1,27 +0,0 @@
-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_details"
-
-[cost]
-input = 0.27
-output = 1.10
-cache_read = 0.548
-
-[limit]
-context = 204800
-output = 131072
-
-[modalities]
-input = ["text"]
-output = ["text"]
-

providers/aihubmix/models/glm-5.1.toml

@@ -1,23 +1,25 @@
 name = "GLM-5.1"
 family = "glm"
-release_date = "2026-04-11"
-last_updated = "2026-04-11"
+release_date = "2026-03-27"
+last_updated = "2026-03-27"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
+structured_output = true
 open_weights = false
 
 [interleaved]
 field = "reasoning_content"
 
 [cost]
-input = 0.84
+input = 0.845
 output = 3.38
+cache_read = 0.183112
 
 [limit]
-context = 200000
-output = 128000
+context = 200_000
+output = 128_000
 
 [modalities]
 input = ["text"]

providers/aihubmix/models/glm-5.toml

@@ -1,24 +0,0 @@
-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/glm-5v-turbo.toml

@@ -0,0 +1,26 @@
+name = "GLM 5 Vision Turbo"
+family = "glm"
+release_date = "2026-05-09"
+last_updated = "2026-05-09"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.7042
+output = 3.09848
+cache_read = 0.169008
+
+[limit]
+context = 200_000
+output = 128_000
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/aihubmix/models/gpt-4.1-nano.toml

@@ -1,23 +0,0 @@
-name = "GPT-4.1 nano"
-family = "gpt-nano"
-release_date = "2025-04-14"
-last_updated = "2025-04-14"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2024-04"
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0.10
-output = 0.40
-cache_read = 0.03
-
-[limit]
-context = 1_047_576
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/aihubmix/models/gpt-4o.toml

@@ -1,23 +0,0 @@
-name = "GPT-4o"
-family = "gpt"
-release_date = "2024-05-13"
-last_updated = "2024-08-06"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2023-09"
-tool_call = true
-open_weights = false
-
-[cost]
-input = 2.50
-output = 10.00
-cache_read = 1.25
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/aihubmix/models/gpt-5-mini.toml

@@ -1,23 +0,0 @@
-name = "GPT-5-Mini"
-family = "gpt-mini"
-release_date = "2025-09-15"
-last_updated = "2025-09-15"
-attachment = true
-reasoning = true
-temperature = true
-knowledge = "2024-09-30"
-tool_call = true
-open_weights = false
-
-[cost]
-input = 1.50
-output = 6.00
-cache_read = 0.75
-
-[limit]
-context = 200_000
-output = 64_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/aihubmix/models/gpt-5-nano.toml

@@ -1,23 +0,0 @@
-name = "GPT-5-Nano"
-family = "gpt-nano"
-release_date = "2025-09-15"
-last_updated = "2025-09-15"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2024-09-30"
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0.50
-output = 2.00
-cache_read = 0.25
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

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

@@ -1,23 +0,0 @@
-name = "GPT-5.1-Codex-Max"
-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
-cache_read = 0.125
-
-[limit]
-context = 400_000
-output = 128_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/aihubmix/models/gpt-5.1-codex-mini.toml

@@ -1,22 +1,24 @@
-name = "GPT-5.1 Codex Mini"
+name = "GPT-5.1 Codex mini"
 family = "gpt-codex"
-release_date = "2025-11-15"
-last_updated = "2025-11-15"
+release_date = "2025-11-13"
+last_updated = "2025-11-13"
 attachment = true
 reasoning = true
-temperature = true
+temperature = false
 tool_call = true
-knowledge = "2025-11"
+structured_output = true
+knowledge = "2024-09-30"
 open_weights = false
 
 [cost]
 input = 0.25
-output = 2.00
-cache_read = 0.03
+output = 2
+cache_read = 0.025
 
 [limit]
 context = 400_000
 output = 128_000
+input = 272_000
 
 [modalities]
 input = ["text", "image"]

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

@@ -1,22 +1,24 @@
 name = "GPT-5.1 Codex"
 family = "gpt-codex"
-release_date = "2025-11-15"
-last_updated = "2025-11-15"
+release_date = "2025-11-13"
+last_updated = "2025-11-13"
 attachment = true
 reasoning = true
-temperature = true
+temperature = false
 tool_call = true
-knowledge = "2025-11"
+structured_output = true
+knowledge = "2024-09-30"
 open_weights = false
 
 [cost]
 input = 1.25
-output = 10.00
-cache_read = 0.13
+output = 10
+cache_read = 0.125
 
 [limit]
 context = 400_000
 output = 128_000
+input = 272_000
 
 [modalities]
 input = ["text", "image"]

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

@@ -0,0 +1,25 @@
+name = "GPT-5.3 Codex"
+family = "gpt-codex"
+release_date = "2026-02-05"
+last_updated = "2026-02-05"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2025-08-31"
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14
+cache_read = 0.175
+
+[limit]
+context = 400_000
+output = 128_000
+input = 272_000
+
+[modalities]
+input = ["text", "image", "pdf"]
+output = ["text"]