fix(vercel): sync audio model types

chore(sync): update OpenRouter model catalog

models/alibaba/qwen3.5-9b.toml

@@ -0,0 +1,22 @@
+name = "Qwen3.5 9B"
+family = "qwen"
+release_date = "2026-02-23"
+last_updated = "2026-02-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/Qwen/Qwen3.5-9B"

models/cohere/command-r7b-12-2024.toml

@@ -1,7 +1,7 @@
 name = "Command R7B"
 family = "command-r"
-release_date = "2024-02-27"
-last_updated = "2024-02-27"
+release_date = "2024-12-02"
+last_updated = "2024-12-02"
 attachment = false
 reasoning = false
 temperature = true

models/google/gemma-4-E2B-it.toml

@@ -0,0 +1,22 @@
+name = "Gemma 4 E2B IT"
+family = "gemma"
+release_date = "2026-04-02"
+last_updated = "2026-04-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[limit]
+context = 131_072
+output = 8_192
+
+[modalities]
+input = ["text", "image", "audio"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/google/gemma-4-E2B-it"

models/google/gemma-4-E4B-it.toml

@@ -0,0 +1,22 @@
+name = "Gemma 4 E4B IT"
+family = "gemma"
+release_date = "2026-04-02"
+last_updated = "2026-04-02"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[limit]
+context = 131_072
+output = 8_192
+
+[modalities]
+input = ["text", "image", "audio"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/google/gemma-4-E4B-it"

models/minimax/MiniMax-M3.toml

@@ -6,7 +6,7 @@ attachment = true
 reasoning = true
 temperature = true
 tool_call = true
-open_weights = false
+open_weights = true
 
 [limit]
 context = 512_000
@@ -15,3 +15,7 @@ output = 128_000
 [modalities]
 input = ["text", "image", "video"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/MiniMaxAI/MiniMax-M3"

models/moonshotai/kimi-k2.7-code-highspeed.toml

@@ -0,0 +1,23 @@
+name = "Kimi K2.7 Code Highspeed"
+family = "kimi-k2"
+release_date = "2026-06-12"
+last_updated = "2026-06-12"
+attachment = true
+reasoning = true
+temperature = false
+tool_call = true
+structured_output = true
+knowledge = "2025-01"
+open_weights = true
+
+[limit]
+context = 262_144
+output = 262_144
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/moonshotai/Kimi-K2.7-Code"

models/openai/gpt-image-1.5.toml

@@ -0,0 +1,17 @@
+name = "GPT-Image-1.5"
+family = "gpt-image"
+release_date = "2025-11-25"
+last_updated = "2025-11-25"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["text", "image"]

models/openai/gpt-image-1.toml

@@ -0,0 +1,17 @@
+name = "GPT-Image-1"
+family = "gpt-image"
+release_date = "2025-04-24"
+last_updated = "2025-04-24"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]

models/openai/gpt-image-2.toml

@@ -0,0 +1,17 @@
+name = "GPT-Image-2"
+family = "gpt-image"
+release_date = "2026-04-21"
+last_updated = "2026-04-21"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]

models/zhipuai/glm-5.1.toml

@@ -1,7 +1,7 @@
 name = "GLM-5.1"
 family = "glm"
-release_date = "2026-03-27"
-last_updated = "2026-03-27"
+release_date = "2026-04-07"
+last_updated = "2026-04-07"
 attachment = false
 reasoning = true
 temperature = true

models/zhipuai/glm-5.2.toml

@@ -7,7 +7,7 @@ reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-open_weights = false
+open_weights = true
 
 [limit]
 context = 1_000_000
@@ -16,3 +16,7 @@ output = 131_072
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/zai-org/GLM-5.2"

models/zhipuai/glm-5.toml

@@ -1,7 +1,7 @@
 name = "GLM-5"
 family = "glm"
-release_date = "2026-02-11"
-last_updated = "2026-02-11"
+release_date = "2026-02-12"
+last_updated = "2026-02-12"
 attachment = false
 reasoning = true
 temperature = true

packages/core/src/family.ts

@@ -246,6 +246,9 @@ export const ModelFamilyValues = [
   // Lucid
   "lucid",
 
+  // LucidQuery
+  "agi",
+
   // Intellect
   "intellect",
 

packages/core/src/sync/providers/openrouter.ts

@@ -7,6 +7,7 @@ import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "
 
 const API_ENDPOINT = "https://openrouter.ai/api/v1/models";
 const MODELS_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "models");
+const MODEL_NAME_BLACKLIST = ["fable-5"];
 const modelMetadataByID = new Map<string, Record<string, unknown>>();
 const modelMetadataFilesByProvider = new Map<string, Set<string>>();
 
@@ -79,7 +80,10 @@ export const openrouter = {
     return response.json();
   },
   parseModels(raw) {
-    return OpenRouterResponse.parse(raw).data;
+    return OpenRouterResponse.parse(raw).data.filter((model) => {
+      const name = `${model.id} ${model.name}`.toLowerCase();
+      return MODEL_NAME_BLACKLIST.every((value) => !name.includes(value));
+    });
   },
   translateModel(model, context) {
     return {

packages/core/src/sync/providers/vercel.ts

@@ -6,7 +6,16 @@ import { factorBaseModel, resolveCanonicalBaseModel } from "./openrouter.js";
 
 const API_ENDPOINT = "https://ai-gateway.vercel.sh/v1/models";
 
-const ModelType = z.enum(["language", "embedding", "image", "video", "reranking"]);
+const ModelType = z.enum([
+  "language",
+  "embedding",
+  "image",
+  "video",
+  "reranking",
+  "transcription",
+  "speech",
+  "realtime",
+]);
 
 const PricingTier = z.object({
   cost: z.string(),
@@ -30,8 +39,8 @@ export const VercelModel = z.object({
   name: z.string(),
   created: z.number(),
   released: z.number().optional(),
-  context_window: z.number(),
-  max_tokens: z.number(),
+  context_window: z.number().optional().default(0),
+  max_tokens: z.number().optional().default(0),
   type: ModelType,
   tags: z.array(z.string()).optional().default([]),
   pricing: Pricing.optional(),
@@ -107,9 +116,17 @@ export function buildVercelModel(model: VercelModel, existing: ExistingModel | u
     cost,
     limit: { context, input, output },
     modalities: {
-      input: ["text", tags.has("vision") ? "image" : undefined, tags.has("file-input") ? "pdf" : undefined]
-        .filter((value): value is "text" | "image" | "pdf" => value !== undefined),
-      output: model.type === "image"
+      input: model.type === "transcription"
+        ? ["audio"]
+        : model.type === "realtime"
+        ? ["text", "audio"]
+        : ["text", tags.has("vision") ? "image" : undefined, tags.has("file-input") ? "pdf" : undefined]
+          .filter((value): value is "text" | "image" | "pdf" => value !== undefined),
+      output: model.type === "speech"
+        ? ["audio"]
+        : model.type === "realtime"
+        ? ["text", "audio"]
+        : model.type === "image"
         ? ["image"]
         : model.type === "video"
         ? ["video"]

packages/core/test/baseten-sync.test.ts

@@ -1,150 +0,0 @@
-import { afterEach, expect, test } from "bun:test";
-import path from "node:path";
-import { mkdir, mkdtemp } from "node:fs/promises";
-import os from "node:os";
-
-import { syncProvider } from "../src/sync/index.js";
-import {
-  BasetenResponse,
-  baseten,
-  buildBasetenModel,
-  fetchBasetenModels,
-  type BasetenModel,
-} from "../src/sync/providers/baseten.js";
-
-const catalogModel: BasetenModel = {
-  id: "zai-org/GLM-5.1",
-  name: "GLM 5.1",
-  context_length: 128_000,
-  max_completion_tokens: 32_000,
-  input_modalities: ["text"],
-  output_modalities: ["text"],
-  pricing: {
-    prompt: "0.00000012",
-    completion: "0.0000005",
-  },
-  supported_features: ["reasoning", "reasoning_effort", "tools", "structured_outputs"],
-  supported_sampling_parameters: ["temperature", "top_p"],
-};
-
-const newCatalogModel: BasetenModel = {
-  ...catalogModel,
-};
-
-afterEach(() => {
-  baseten.modelsDir = "providers/baseten/models";
-  baseten.fetchModels = async () => {
-    const key = process.env.BASETEN_API_KEY;
-    if (key === undefined) throw new Error("Baseten sync requires BASETEN_API_KEY");
-    return fetchBasetenModels(key);
-  };
-});
-
-test("Baseten maps authoritative fields and preserves curated metadata", () => {
-  const synced = buildBasetenModel(catalogModel, {
-    name: "Old name",
-    release_date: "2025-08-05",
-    last_updated: "2025-09-01",
-    attachment: false,
-    reasoning: true,
-    reasoning_options: [{ type: "effort", values: ["low", "high"] }],
-    tool_call: true,
-    open_weights: true,
-    status: "deprecated",
-    interleaved: { field: "reasoning_content" },
-    base_model: "zhipuai/glm-5.1",
-    base_model_omit: ["limit.input"],
-    cost: { input: 0.1, output: 0.4, cache_write: 0.2 },
-    limit: { context: 64_000, output: 16_000 },
-    modalities: { input: ["text"], output: ["text"] },
-  });
-
-  expect(synced).toMatchObject({
-    base_model: "zhipuai/glm-5.1",
-    base_model_omit: ["limit.input"],
-    reasoning_options: [{ type: "effort", values: ["low", "high"] }],
-    status: "deprecated",
-    interleaved: { field: "reasoning_content" },
-    cost: { input: 0.12, output: 0.5, cache_write: 0.2 },
-    limit: { context: 128_000, output: 32_000 },
-  });
-});
-
-test("Baseten preserves curated reasoning when an opt-in capability is omitted", () => {
-  const synced = buildBasetenModel({
-    ...catalogModel,
-    supported_features: ["tools", "structured_outputs"],
-  }, {
-    name: "GLM 5.1",
-    release_date: "2026-05-20",
-    last_updated: "2026-05-20",
-    attachment: false,
-    reasoning: true,
-    reasoning_options: [{ type: "toggle" }],
-    tool_call: true,
-    open_weights: true,
-    cost: { input: 1, output: 4 },
-    limit: { context: 100_000, output: 50_000 },
-    modalities: { input: ["text"], output: ["text"] },
-  });
-
-  expect(synced).toMatchObject({
-    reasoning: true,
-    reasoning_options: [{ type: "toggle" }],
-  });
-});
-
-test("Baseten sync adds exact base models, retains missing entries, and is idempotent", async () => {
-  const root = await mkdtemp(path.join(os.tmpdir(), "models-dev-baseten-"));
-  const modelsDir = path.join(root, "providers", "baseten", "models");
-  const metadataDir = path.join(root, "models", "zhipuai");
-  await mkdir(path.join(modelsDir, "stale"), { recursive: true });
-  await mkdir(metadataDir, { recursive: true });
-  await Bun.write(
-    path.join(metadataDir, "glm-5.1.toml"),
-    Bun.file(path.join(import.meta.dirname, "../../../models/zhipuai/glm-5.1.toml")),
-  );
-  await Bun.write(path.join(modelsDir, "stale", "model.toml"), [
-    'name = "Retained"',
-    'release_date = "2025-01-01"',
-    'last_updated = "2025-01-01"',
-    "attachment = false",
-    "reasoning = false",
-    "tool_call = false",
-    "open_weights = false",
-    "[cost]",
-    "input = 1",
-    "output = 1",
-    "[limit]",
-    "context = 1000",
-    "output = 100",
-    "[modalities]",
-    'input = ["text"]',
-    'output = ["text"]',
-    "",
-  ].join("\n"));
-  baseten.modelsDir = modelsDir;
-  baseten.fetchModels = async () => ({ data: [newCatalogModel] });
-
-  const first = await syncProvider(baseten);
-  const second = await syncProvider(baseten);
-
-  expect(first.created).toBe(1);
-  expect(first.deleted).toBe(0);
-  expect(first.notices.join(" ")).toContain("stale/model.toml");
-  expect(second).toMatchObject({ created: 0, updated: 0, deleted: 0 });
-});
-
-test("Baseten rejects malformed catalog responses", () => {
-  expect(() => BasetenResponse.parse({ data: "broken" })).toThrow();
-});
-
-test("Baseten rejects non-success API responses", async () => {
-  const fetcher = async () => new Response("unauthorized", {
-    status: 401,
-    statusText: "Unauthorized",
-  });
-
-  expect(fetchBasetenModels("fixture-key", fetcher as typeof fetch))
-    .rejects.toThrow("Baseten models request failed: 401 Unauthorized");
-});

packages/core/test/cloudflare-workers-ai-sync.test.ts

@@ -1,35 +0,0 @@
-import { expect, test } from "bun:test";
-
-import { buildWorkersAiModel } from "../src/sync/providers/cloudflare-workers-ai.js";
-import type { OpenRouterModel } from "../src/sync/providers/openrouter.js";
-
-test("Cloudflare Workers AI sync preserves reasoning options", () => {
-  const model: OpenRouterModel = {
-    id: "@cf/nvidia/nemotron-3-120b-a12b",
-    name: "Nemotron 3 Super 120B",
-    created: 1_773_187_200,
-    hugging_face_id: "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
-    knowledge_cutoff: null,
-    context_length: 256_000,
-    architecture: {
-      input_modalities: ["text"],
-      output_modalities: ["text"],
-    },
-    pricing: {
-      prompt: "0.0000005",
-      completion: "0.0000015",
-    },
-    top_provider: {
-      context_length: 256_000,
-      max_completion_tokens: 256_000,
-    },
-    supported_parameters: ["reasoning", "tools", "temperature"],
-  };
-
-  const synced = buildWorkersAiModel(model, {
-    base_model: "nvidia/nemotron-3-super-120b-a12b",
-    reasoning_options: [{ type: "toggle" }],
-  });
-
-  expect(synced.reasoning_options).toEqual([{ type: "toggle" }]);
-});

packages/core/test/google-sync.test.ts

@@ -1,35 +0,0 @@
-import { expect, test } from "bun:test";
-
-import { buildGoogleModel } from "../src/sync/providers/google.js";
-
-test("Google sync keeps base models compact", () => {
-  const synced = buildGoogleModel({
-    name: "models/gemini-3-pro-image-preview",
-    displayName: "Nano Banana Pro",
-    inputTokenLimit: 131_072,
-    outputTokenLimit: 32_768,
-    temperature: 1,
-    thinking: true,
-  }, {
-    base_model: "google/gemini-3-pro-image-preview",
-    name: "Nano Banana Pro",
-    family: "gemini-pro",
-    release_date: "2025-11-20",
-    last_updated: "2025-11-20",
-    attachment: true,
-    reasoning: true,
-    temperature: true,
-    tool_call: false,
-    knowledge: "2025-01",
-    open_weights: false,
-    cost: { input: 2, output: 120 },
-    limit: { context: 65_536, output: 32_768 },
-    modalities: { input: ["text", "image"], output: ["text", "image"] },
-  });
-
-  expect(synced).toEqual({
-    base_model: "google/gemini-3-pro-image-preview",
-    cost: { input: 2, output: 120 },
-    limit: { context: 131_072 },
-  });
-});

packages/core/test/openrouter-sync.test.ts

@@ -1,121 +0,0 @@
-import { expect, test } from "bun:test";
-
-import { preserveBaseModel } from "../src/sync/index.js";
-import { resolveCloudflareBaseModel } from "../src/sync/providers/cloudflare-workers-ai.js";
-import { buildOpenRouterModel, type OpenRouterModel } from "../src/sync/providers/openrouter.js";
-
-test("OpenRouter z-ai models inherit from zhipuai metadata", () => {
-  const model: OpenRouterModel = {
-    id: "z-ai/glm-5.1",
-    name: "Z.AI: GLM-5.1",
-    created: 1_777_680_000,
-    hugging_face_id: "zai-org/GLM-5.1",
-    knowledge_cutoff: null,
-    context_length: 200_000,
-    architecture: {
-      input_modalities: ["text"],
-      output_modalities: ["text"],
-    },
-    pricing: {
-      prompt: "0.0000014",
-      completion: "0.0000044",
-    },
-    top_provider: {
-      context_length: 200_000,
-      max_completion_tokens: 131_072,
-    },
-    supported_parameters: ["tools", "tool_choice", "temperature", "structured_outputs"],
-  };
-
-  const synced = buildOpenRouterModel(model, undefined);
-
-  expect("base_model" in synced ? synced.base_model : undefined).toBe("zhipuai/glm-5.1");
-});
-
-test("OpenRouter-derived syncs preserve existing base model links", () => {
-  const model: OpenRouterModel = {
-    id: "@cf/nvidia/nemotron-3-120b-a12b",
-    name: "Nemotron 3 Super 120B",
-    created: 1_773_187_200,
-    hugging_face_id: "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
-    knowledge_cutoff: null,
-    context_length: 256_000,
-    architecture: {
-      input_modalities: ["text"],
-      output_modalities: ["text"],
-    },
-    pricing: {
-      prompt: "0.0000005",
-      completion: "0.0000015",
-    },
-    top_provider: {
-      context_length: 256_000,
-      max_completion_tokens: 256_000,
-    },
-    supported_parameters: ["reasoning", "tools", "temperature", "structured_outputs"],
-  };
-
-  const synced = preserveBaseModel(buildOpenRouterModel(model, undefined), {
-    base_model: "nvidia/nemotron-3-super-120b-a12b",
-    base_model_omit: ["limit.input"],
-  });
-
-  expect("base_model" in synced ? synced.base_model : undefined)
-    .toBe("nvidia/nemotron-3-super-120b-a12b");
-  expect("base_model_omit" in synced ? synced.base_model_omit : undefined)
-    .toEqual(["limit.input"]);
-});
-
-test("newly detected base models do not replace existing links", () => {
-  const model: OpenRouterModel = {
-    id: "z-ai/glm-5.1",
-    name: "Z.AI: GLM-5.1",
-    created: 1_777_680_000,
-    hugging_face_id: null,
-    knowledge_cutoff: null,
-    context_length: 200_000,
-    architecture: { input_modalities: ["text"], output_modalities: ["text"] },
-    pricing: { prompt: "0.0000014", completion: "0.0000044" },
-    top_provider: { context_length: 200_000, max_completion_tokens: 131_072 },
-    supported_parameters: ["tools"],
-  };
-  const synced = buildOpenRouterModel(model, {
-    base_model: "zhipuai/glm-5",
-  });
-
-  expect("base_model" in synced ? synced.base_model : undefined).toBe("zhipuai/glm-5");
-});
-
-test("undefined translated links preserve existing base model fields", () => {
-  const synced = preserveBaseModel({
-    base_model: undefined,
-  } as never, {
-    base_model: "nvidia/nemotron-3-super-120b-a12b",
-    base_model_omit: ["limit.input"],
-  });
-
-  expect("base_model" in synced ? synced.base_model : undefined)
-    .toBe("nvidia/nemotron-3-super-120b-a12b");
-  expect("base_model_omit" in synced ? synced.base_model_omit : undefined)
-    .toEqual(["limit.input"]);
-});
-
-test("new Cloudflare models discover a unique metadata base model", () => {
-  const model: OpenRouterModel = {
-    id: "@cf/nvidia/nemotron-3-120b-a12b",
-    name: "Nemotron 3 Super 120B",
-    created: 1_773_187_200,
-    hugging_face_id: null,
-    knowledge_cutoff: null,
-    context_length: 256_000,
-    architecture: { input_modalities: ["text"], output_modalities: ["text"] },
-    pricing: { prompt: "0.0000005", completion: "0.0000015" },
-    top_provider: { context_length: 256_000, max_completion_tokens: 256_000 },
-    supported_parameters: ["reasoning"],
-  };
-
-  expect(resolveCloudflareBaseModel(model)).toBe("nvidia/nemotron-3-super-120b-a12b");
-  const synced = buildOpenRouterModel(model, undefined, resolveCloudflareBaseModel(model));
-  expect("base_model" in synced ? synced.base_model : undefined)
-    .toBe("nvidia/nemotron-3-super-120b-a12b");
-});

packages/core/test/sync-runner.test.ts

@@ -1,227 +0,0 @@
-import { expect, test } from "bun:test";
-import path from "node:path";
-import { mkdtemp, mkdir, readlink, symlink } from "node:fs/promises";
-import os from "node:os";
-
-import { AuthoredModelShape } from "../src/schema.js";
-import { syncProvider, type SyncProvider, type SyncedFullModel } from "../src/sync/index.js";
-
-const model: SyncedFullModel = {
-  name: "Test model",
-  release_date: "2026-01-01",
-  last_updated: "2026-01-01",
-  attachment: false,
-  reasoning: false,
-  tool_call: false,
-  open_weights: false,
-  cost: { input: 1, output: 2 },
-  limit: { context: 1_000, output: 100 },
-  modalities: { input: ["text"], output: ["text"] },
-};
-
-test("reasoning budgets allow only the -1 negative sentinel", () => {
-  const authored = { id: "model", ...model };
-
-  expect(AuthoredModelShape.safeParse({
-    ...authored,
-    reasoning_options: [{ type: "budget_tokens", min: -1, max: 32_768 }],
-  }).success).toBe(true);
-  expect(AuthoredModelShape.safeParse({
-    ...authored,
-    reasoning_options: [{ type: "budget_tokens", min: -2, max: 32_768 }],
-  }).success).toBe(false);
-});
-
-test("reasoning efforts accept the provider default value", () => {
-  expect(AuthoredModelShape.safeParse({
-    id: "model",
-    ...model,
-    reasoning: true,
-    reasoning_options: [{ type: "effort", values: ["none", "default"] }],
-  }).success).toBe(true);
-});
-
-async function fixture() {
-  const root = await mkdtemp(path.join(os.tmpdir(), "models-dev-sync-"));
-  const modelsDir = path.join(root, "providers", "test", "models");
-  await mkdir(modelsDir, { recursive: true });
-  return { root, modelsDir };
-}
-
-function provider(
-  modelsDir: string,
-  ids: string[],
-  deleteMissing = true,
-  preserveSymlinks = false,
-): SyncProvider<string> {
-  return {
-    id: "test",
-    name: "Test",
-    modelsDir,
-    deleteMissing,
-    preserveSymlinks,
-    missingNotice: (paths) => paths.map((item) => `missing: ${item}`),
-    async fetchModels() {
-      return ids;
-    },
-    parseModels(raw) {
-      return raw as string[];
-    },
-    translateModel(id) {
-      return { id, model };
-    },
-  };
-}
-
-test("sync repairs a broken symlink returned by the source", async () => {
-  const { modelsDir } = await fixture();
-  const filePath = path.join(modelsDir, "model.toml");
-  await symlink("missing.toml", filePath);
-
-  const result = await syncProvider(provider(modelsDir, ["model"]));
-
-  expect(result.created).toBe(1);
-  expect(await Bun.file(filePath).text()).toContain('name = "Test model"');
-  expect(readlink(filePath)).rejects.toThrow();
-});
-
-test("sync preserves valid symlink aliases when configured", async () => {
-  const { root, modelsDir } = await fixture();
-  const targetPath = path.join(root, "target.toml");
-  const filePath = path.join(modelsDir, "model.toml");
-  await Bun.write(targetPath, `name = "Alias target"\n`);
-  await symlink(targetPath, filePath);
-
-  const result = await syncProvider(provider(modelsDir, ["model"], true, true));
-
-  expect(result.updated).toBe(0);
-  expect(await readlink(filePath)).toBe(targetPath);
-});
-
-test("sync removes a broken symlink absent from the source", async () => {
-  const { modelsDir } = await fixture();
-  const filePath = path.join(modelsDir, "model.toml");
-  await symlink("missing.toml", filePath);
-
-  const result = await syncProvider(provider(modelsDir, []));
-
-  expect(result.deleted).toBe(1);
-  expect(await Bun.file(filePath).exists()).toBe(false);
-});
-
-test("non-deleting sync reports missing broken symlinks", async () => {
-  const { modelsDir } = await fixture();
-  const filePath = path.join(modelsDir, "model.toml");
-  await symlink("missing.toml", filePath);
-
-  const result = await syncProvider(provider(modelsDir, [], false));
-
-  expect(result.deleted).toBe(0);
-  expect(result.notices).toEqual(["missing: model.toml"]);
-  expect(await readlink(filePath)).toBe("missing.toml");
-});
-
-test("sync preserves authored reasoning options omitted by a translator", async () => {
-  const { modelsDir } = await fixture();
-  const filePath = path.join(modelsDir, "model.toml");
-  await Bun.write(filePath, `name = "Old name"
-release_date = "2026-01-01"
-last_updated = "2026-01-01"
-attachment = false
-reasoning = true
-tool_call = false
-open_weights = false
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high"]
-
-[[reasoning_options]]
-type = "budget_tokens"
-min = -1
-max = 32768
-
-[cost]
-input = 1
-output = 2
-
-[limit]
-context = 1000
-output = 100
-
-[modalities]
-input = ["text"]
-output = ["text"]
-`);
-  const sync = provider(modelsDir, ["model"]);
-  sync.translateModel = (id) => ({
-    id,
-    model: { ...model, reasoning: true },
-  });
-
-  const first = await syncProvider(sync);
-  const content = await Bun.file(filePath).text();
-  const second = await syncProvider(sync);
-
-  expect(first.updated).toBe(1);
-  expect(content).toContain("[[reasoning_options]]");
-  expect(content).toContain('values = ["low", "high"]');
-  expect(content).toContain("min = -1");
-  expect(content).toContain("max = 32_768");
-  expect(second.updated).toBe(0);
-  expect(second.unchanged).toBe(1);
-});
-
-test("sync writes metadata returned by a provider translator", async () => {
-  const root = await mkdtemp(path.join(os.tmpdir(), "models-dev-sync-metadata-"));
-  const modelsDir = path.join(root, "providers", "test", "models");
-  await mkdir(modelsDir, { recursive: true });
-  const sync = provider(modelsDir, ["model"]);
-  sync.translateModel = () => ({
-    id: "model",
-    model: {
-      base_model: "test/model",
-      reasoning_options: [],
-      cost: { input: 1, output: 2 },
-    },
-    metadata: {
-      id: "test/model",
-      model: {
-        name: "Model",
-        release_date: "2026-06-10",
-        last_updated: "2026-06-10",
-        attachment: false,
-        reasoning: false,
-        tool_call: true,
-        open_weights: false,
-        limit: { context: 1_000, output: 100 },
-        modalities: { input: ["text"], output: ["text"] },
-      },
-    },
-  });
-
-  const first = await syncProvider(sync);
-  const second = await syncProvider(sync);
-
-  expect(first).toMatchObject({ created: 2, updated: 0 });
-  expect(second).toMatchObject({ created: 0, updated: 0 });
-  expect(await Bun.file(path.join(root, "models", "test", "model.toml")).text()).toContain('name = "Model"');
-});
-
-test("sync removes missing metadata only from its owned namespace", async () => {
-  const { root, modelsDir } = await fixture();
-  const ownedDir = path.join(root, "models", "test");
-  const otherDir = path.join(root, "models", "other");
-  await mkdir(ownedDir, { recursive: true });
-  await mkdir(otherDir, { recursive: true });
-  await Bun.write(path.join(ownedDir, "stale.toml"), 'name = "Stale"\n');
-  await Bun.write(path.join(otherDir, "retained.toml"), 'name = "Retained"\n');
-  const sync = provider(modelsDir, []);
-  sync.metadataNamespace = "test";
-
-  const result = await syncProvider(sync);
-
-  expect(result.deleted).toBe(1);
-  expect(await Bun.file(path.join(ownedDir, "stale.toml")).exists()).toBe(false);
-  expect(await Bun.file(path.join(otherDir, "retained.toml")).exists()).toBe(true);
-});

packages/core/test/venice-sync.test.ts

@@ -1,167 +0,0 @@
-import { expect, test } from "bun:test";
-import { readdirSync } from "node:fs";
-import path from "node:path";
-
-import {
-  buildVeniceModel,
-  resolveVeniceBaseModel,
-  venice,
-  VeniceResponse,
-  type VeniceModel,
-} from "../src/sync/providers/venice.js";
-
-const catalogModel: VeniceModel = {
-  id: "openai-gpt-54",
-  created: 1_772_668_800,
-  model_spec: {
-    name: "GPT-5.4",
-    availableContextTokens: 400_000,
-    maxCompletionTokens: 128_000,
-    modelSource: "OpenAI",
-    capabilities: {
-      supportsVision: true,
-      supportsReasoning: true,
-      supportsReasoningEffort: true,
-      reasoningEffortOptions: ["none", "low", "medium", "high"],
-      supportsFunctionCalling: true,
-      supportsResponseSchema: true,
-    },
-    pricing: {
-      input: { usd: 3.13 },
-      output: { usd: 18.75 },
-      cache_input: { usd: 0.313 },
-      extended: {
-        context_token_threshold: 200_000,
-        input: { usd: 6.26 },
-        output: { usd: 28.125 },
-      },
-    },
-  },
-};
-
-test("Venice resolves flattened IDs to canonical metadata", () => {
-  expect(resolveVeniceBaseModel("openai-gpt-54", "GPT-5.4")).toBe("openai/gpt-5.4");
-  expect(resolveVeniceBaseModel("claude-opus-4-8-fast", "Claude Opus 4.8 Fast"))
-    .toBe("anthropic/claude-opus-4-8");
-});
-
-test("Venice emits empty reasoning options when efforts are unavailable", () => {
-  const synced = buildVeniceModel({
-    ...catalogModel,
-    id: "reasoning-without-efforts",
-    model_spec: {
-      ...catalogModel.model_spec,
-      name: "Reasoning Without Efforts",
-      capabilities: {
-        ...catalogModel.model_spec.capabilities,
-        reasoningEffortOptions: [],
-      },
-    },
-  }, undefined, undefined, "2026-06-10");
-
-  expect(synced).toMatchObject({ reasoning: true, reasoning_options: [] });
-});
-
-test("Venice does not infer temperature support", () => {
-  const synced = buildVeniceModel(catalogModel, undefined, null, "2026-06-10");
-
-  expect(synced.temperature).toBeUndefined();
-});
-
-test("Venice skips E2EE models", () => {
-  const translated = venice.translateModel({
-    ...catalogModel,
-    id: "e2ee-test-model",
-    model_spec: {
-      ...catalogModel.model_spec,
-      capabilities: { ...catalogModel.model_spec.capabilities, supportsE2EE: true },
-    },
-  }, { existing: () => undefined });
-
-  expect(translated).toBeUndefined();
-});
-
-test("Venice uses boundary-aware family matching", () => {
-  const synced = buildVeniceModel({
-    ...catalogModel,
-    id: "google-gemma-4-31b-it",
-    model_spec: { ...catalogModel.model_spec, name: "Google Gemma 4 31B Instruct" },
-  }, undefined, null, "2026-06-10");
-
-  expect(synced).toMatchObject({ family: "gemma" });
-});
-
-test("Venice maps API fields without bumping inherited model timestamps", () => {
-  const synced = buildVeniceModel(catalogModel, {
-    base_model: "openai/gpt-5.4",
-    name: "GPT-5.4",
-    family: "gpt",
-    release_date: "2026-03-05",
-    last_updated: "2026-03-09",
-    attachment: true,
-    reasoning: true,
-    tool_call: true,
-    structured_output: true,
-    temperature: true,
-    open_weights: false,
-    interleaved: { field: "reasoning_content" },
-    cost: { input: 3, output: 18, input_audio: 4 },
-    limit: { context: 400_000, output: 128_000 },
-    modalities: { input: ["text", "image", "pdf"], output: ["text"] },
-  }, "openai/gpt-5.4", "2026-06-10");
-
-  expect(synced).toMatchObject({
-    base_model: "openai/gpt-5.4",
-    base_model_omit: ["limit.input"],
-    last_updated: "2026-03-09",
-    reasoning_options: [{ type: "effort", values: ["none", "low", "medium", "high"] }],
-    interleaved: { field: "reasoning_content" },
-    cost: {
-      input: 3.13,
-      output: 18.75,
-      cache_read: 0.313,
-      input_audio: 4,
-      tiers: [{ tier: { type: "context", size: 200_000 }, input: 6.26, output: 28.125 }],
-    },
-  });
-  expect(synced).not.toHaveProperty("family");
-  expect(synced).not.toHaveProperty("release_date");
-  expect(synced).not.toHaveProperty("open_weights");
-  expect(synced).not.toHaveProperty("modalities");
-  expect(synced).not.toHaveProperty("temperature");
-});
-
-test("Venice preserves last_updated when authoritative data is unchanged", () => {
-  const providerModel = {
-    ...catalogModel,
-    id: "venice-only-test-model",
-    model_spec: { ...catalogModel.model_spec, name: "Venice Only Test Model" },
-  };
-  const full = buildVeniceModel(providerModel, undefined, undefined, "2026-06-10");
-  if ("base_model" in full) throw new Error("Expected a full provider model fixture");
-  const synced = buildVeniceModel(providerModel, full, undefined, "2026-06-11");
-
-  expect(synced).toMatchObject({ last_updated: "2026-06-10" });
-});
-
-test("Venice rejects malformed responses", () => {
-  expect(() => VeniceResponse.parse({ data: [{ id: "broken" }] })).toThrow();
-});
-
-test("Venice models use only canonical metadata and declare reasoning options", async () => {
-  const root = path.join(import.meta.dirname, "..", "..", "..");
-  const modelsDir = path.join(root, "providers", "venice", "models");
-
-  for (const file of readdirSync(modelsDir).filter((item) => item.endsWith(".toml"))) {
-    const model = Bun.TOML.parse(await Bun.file(path.join(modelsDir, file)).text()) as {
-      base_model?: string;
-      reasoning_options?: unknown[];
-    };
-    expect(model.reasoning_options, file).toBeDefined();
-    if (model.base_model !== undefined) {
-      expect(model.base_model.startsWith("venice/"), file).toBe(false);
-      expect(await Bun.file(path.join(root, "models", `${model.base_model}.toml`)).exists(), file).toBe(true);
-    }
-    expect(file.startsWith("e2ee-"), file).toBe(false);
-  }
-});

packages/core/test/vercel-sync.test.ts

@@ -1,113 +0,0 @@
-import { expect, test } from "bun:test";
-
-import { buildVercelModel, type VercelModel, vercel } from "../src/sync/providers/vercel.js";
-
-const model: VercelModel = {
-  id: "openai/gpt-test",
-  name: "GPT Test",
-  created: 1_700_000_000,
-  released: 1_710_000_000,
-  context_window: 128_000,
-  max_tokens: 32_000,
-  type: "language",
-  tags: ["reasoning", "tool-use", "vision", "file-input"],
-  pricing: {
-    input: "0.000001",
-    output: "0.000004",
-    input_cache_read: "0.0000001",
-  },
-};
-
-test("Vercel models translate gateway metadata", () => {
-  const synced = buildVercelModel(model, undefined);
-
-  expect(synced).toMatchObject({
-    name: "GPT Test",
-    release_date: "2024-03-09",
-    last_updated: "2024-03-09",
-    attachment: true,
-    reasoning: true,
-    tool_call: true,
-    open_weights: false,
-    cost: { input: 1, output: 4, cache_read: 0.1 },
-    limit: { context: 128_000, input: 96_000, output: 32_000 },
-    modalities: { input: ["text", "image", "pdf"], output: ["text"] },
-  });
-});
-
-test("Vercel models preserve curated metadata and missing limits", () => {
-  const synced = buildVercelModel({
-    ...model,
-    context_window: 0,
-    max_tokens: 0,
-  }, {
-    name: "Curated name",
-    release_date: "2024-01-01",
-    last_updated: "2025-01-01",
-    reasoning_options: [{ type: "effort", values: ["low", "high"] }],
-    cost: {
-      input: 2,
-      output: 8,
-      tiers: [{
-        tier: { type: "context", size: 200_000 },
-        input: 3,
-        output: 12,
-      }],
-    },
-    limit: { context: 64_000, input: 48_000, output: 16_000 },
-  });
-
-  expect(synced.name).toBe("Curated name");
-  expect(synced.last_updated).toBe("2025-01-01");
-  expect(synced.reasoning_options).toEqual([{ type: "effort", values: ["low", "high"] }]);
-  expect(synced.cost?.tiers).toHaveLength(1);
-  expect(synced.limit).toEqual({ context: 64_000, input: 48_000, output: 16_000 });
-});
-
-test("Vercel non-language models use API tool capabilities", () => {
-  const synced = buildVercelModel({
-    ...model,
-    type: "image",
-    tags: [],
-  }, {
-    tool_call: true,
-  });
-
-  expect(synced.tool_call).toBe(false);
-});
-
-test("Vercel sync includes non-language model types", () => {
-  for (const [type, output] of [
-    ["image", ["image"]],
-    ["video", ["video"]],
-    ["reranking", ["text"]],
-  ] as const) {
-    const source = {
-      ...model,
-      id: `test/${type}`,
-      type,
-      tags: [],
-      context_window: 0,
-      max_tokens: 0,
-      pricing: undefined,
-    };
-
-    expect(vercel.translateModel(source, { existing: () => undefined })).toBeDefined();
-    expect(buildVercelModel(source, undefined)).toMatchObject({
-      tool_call: false,
-      modalities: { input: ["text"], output },
-    });
-  }
-});
-
-test("Vercel models use canonical metadata when available", () => {
-  const synced = buildVercelModel({
-    ...model,
-    id: "nvidia/nemotron-3-ultra-550b-a55b",
-    name: "Nemotron 3 Ultra",
-  }, undefined);
-
-  expect("base_model" in synced ? synced.base_model : undefined)
-    .toBe("nvidia/nemotron-3-ultra-550b-a55b");
-  expect("last_updated" in synced ? synced.last_updated : undefined).toBeUndefined();
-});

packages/core/test/xai-sync.test.ts

@@ -1,30 +0,0 @@
-import { expect, test } from "bun:test";
-
-import { buildXAIModel, type XAIModel } from "../src/sync/providers/xai.js";
-
-const model: XAIModel = {
-  id: "grok-test",
-  created: 1_700_000_000,
-  input_modalities: ["text"],
-  output_modalities: ["text"],
-  prompt_text_token_price: 10_000,
-  completion_text_token_price: 20_000,
-};
-
-test("xAI sync preserves reasoning options", () => {
-  const synced = buildXAIModel(model, {
-    name: "Grok Test",
-    release_date: "2024-01-01",
-    last_updated: "2024-01-01",
-    attachment: false,
-    reasoning: true,
-    reasoning_options: [{ type: "effort", values: ["none", "high"] }],
-    tool_call: true,
-    open_weights: false,
-    limit: { context: 2_000_000, output: 30_000 },
-    modalities: { input: ["text"], output: ["text"] },
-  });
-
-  expect(synced.reasoning_options).toEqual([{ type: "effort", values: ["none", "high"] }]);
-  expect(synced.limit?.context).toBe(2_000_000);
-});

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

@@ -3,6 +3,7 @@ release_date = "2026-02-18"
 last_updated = "2026-03-13"
 attachment = true
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 open_weights = false

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

@@ -4,6 +4,7 @@ release_date = "2026-02-18"
 last_updated = "2026-03-13"
 attachment = true
 reasoning = true
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens", min = 1024, max = 63999 }]
 temperature = true
 tool_call = true
 open_weights = false

providers/302ai/models/deepseek-reasoner.toml

@@ -4,6 +4,7 @@ release_date = "2025-01-20"
 last_updated = "2025-01-20"
 attachment = false
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 open_weights = false

providers/302ai/models/deepseek-v3.2-thinking.toml

@@ -3,6 +3,7 @@ release_date = "2025-12-01"
 last_updated = "2025-12-01"
 attachment = false
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 open_weights = false

providers/302ai/models/doubao-seed-1-6-thinking-250715.toml

@@ -3,6 +3,7 @@ release_date = "2025-07-15"
 last_updated = "2025-07-15"
 attachment = true
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 open_weights = false

providers/302ai/models/kimi-k2-thinking-turbo.toml

@@ -3,6 +3,7 @@ release_date = "2025-09-05"
 last_updated = "2025-09-05"
 attachment = false
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 open_weights = false

providers/302ai/models/kimi-k2-thinking.toml

@@ -3,6 +3,7 @@ release_date = "2025-09-05"
 last_updated = "2025-09-05"
 attachment = false
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 open_weights = false

providers/aihubmix/models/grok-4.3.toml

@@ -4,6 +4,7 @@ release_date = "2026-05-01"
 last_updated = "2026-05-01"
 attachment = true
 reasoning = true
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 temperature = true
 tool_call = true
 structured_output = true

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

@@ -4,6 +4,7 @@ release_date = "2026-01"
 last_updated = "2026-01"
 attachment = true
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = false
 tool_call = true
 structured_output = true

providers/aihubmix/models/kimi-k2.6.toml

@@ -4,6 +4,7 @@ release_date = "2026-04-21"
 last_updated = "2026-04-21"
 attachment = true
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = false
 tool_call = true
 structured_output = true

providers/aihubmix/models/qwen3.6-flash.toml

@@ -4,6 +4,7 @@ release_date = "2026-04-02"
 last_updated = "2026-04-02"
 attachment = true
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 structured_output = true

providers/aihubmix/models/qwen3.6-max-preview.toml

@@ -4,6 +4,7 @@ release_date = "2026-05-09"
 last_updated = "2026-05-09"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 structured_output = true

providers/aihubmix/models/qwen3.6-plus.toml

@@ -4,6 +4,7 @@ release_date = "2026-05-09"
 last_updated = "2026-05-09"
 attachment = true
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 structured_output = true

providers/alibaba-token-plan-cn/models/glm-5.2.toml

@@ -0,0 +1,14 @@
+base_model = "zhipuai/glm-5.2"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

providers/alibaba-token-plan-cn/models/kimi-k2.7-code.toml

@@ -0,0 +1,13 @@
+base_model = "moonshotai/kimi-k2.7-code"
+base_model_omit = ["structured_output"]
+family = "kimi-k2"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

providers/alibaba-token-plan/models/glm-5.2.toml

@@ -0,0 +1,14 @@
+base_model = "zhipuai/glm-5.2"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

providers/alibaba-token-plan/models/kimi-k2.7-code.toml

@@ -0,0 +1,13 @@
+base_model = "moonshotai/kimi-k2.7-code"
+base_model_omit = ["structured_output"]
+family = "kimi-k2"
+reasoning_options = [{ type = "toggle" }, { type = "budget_tokens" }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0
+output = 0
+cache_read = 0
+cache_write = 0

providers/ambient/models/moonshotai/kimi-k2.7-code.toml

@@ -0,0 +1,8 @@
+base_model = "moonshotai/kimi-k2.7-code"
+temperature = true
+
+[cost]
+input = 0.75
+output = 3.5
+cache_read = 0.16
+cache_write = 0

providers/ambient/models/zai-org/GLM-5.1-FP8.toml

@@ -1,5 +1,6 @@
-reasoning_options = []
 base_model = "zhipuai/glm-5.1"
+name = "GLM 5.1"
+reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"

providers/azure-cognitive-services/models/kimi-k2.5.toml

@@ -1 +0,0 @@
-../../azure/models/kimi-k2.5.toml

providers/azure-cognitive-services/models/kimi-k2.5.toml

@@ -0,0 +1,16 @@
+base_model = "moonshotai/kimi-k2.5"
+reasoning_options = [{ type = "toggle" }]
+temperature = true
+interleaved = true
+
+[cost]
+input = 0.60
+output = 3.00
+
+[modalities]
+input = ["text", "image"]
+
+[provider]
+shape = "completions"
+npm = "@ai-sdk/openai-compatible"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/models"

providers/azure-cognitive-services/models/kimi-k2.6.toml

@@ -1,30 +1,16 @@
-name = "Kimi K2.6"
-family = "kimi-k2"
-release_date = "2026-04-22"
-last_updated = "2026-04-22"
+base_model = "moonshotai/kimi-k2.6"
 attachment = false
-reasoning = true
 reasoning_options = [{ type = "toggle" }]
-structured_output = true
-temperature = true
-tool_call = true
-knowledge = "2025-01"
 interleaved = true
-open_weights = true
 
 [cost]
 input = 0.95
 output = 4.00
 
-[limit]
-context = 262_144
-output = 262_144
-
 [modalities]
 input = ["text", "image"]
-output = ["text"]
 
 [provider]
 shape = "completions"
 npm = "@ai-sdk/openai-compatible"
-api = "https://${AZURE_RESOURCE_NAME}.services.ai.azure.com/models"
+api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/models"

providers/azure/models/gpt-image-1.5.toml

@@ -0,0 +1,7 @@
+base_model = "openai/gpt-image-1.5"
+reasoning_options = []
+
+[cost]
+input = 5.00
+output = 32.00
+cache_read = 1.25

providers/azure/models/gpt-image-1.toml

@@ -0,0 +1,7 @@
+base_model = "openai/gpt-image-1"
+reasoning_options = []
+
+[cost]
+input = 5.00
+output = 40.00
+cache_read = 1.25

providers/azure/models/gpt-image-2.toml

@@ -0,0 +1,7 @@
+base_model = "openai/gpt-image-2"
+reasoning_options = []
+
+[cost]
+input = 5.00
+output = 30.00
+cache_read = 1.25

providers/baseten/models/moonshotai/Kimi-K2.7-Code.toml

@@ -0,0 +1,14 @@
+base_model = "moonshotai/kimi-k2.7-code"
+temperature = true
+
+[cost]
+input = 0.95
+output = 4
+cache_read = 0.16
+
+[limit]
+context = 262_000
+output = 262_000
+
+[modalities]
+input = ["text", "image"]

providers/baseten/models/nvidia/Nemotron-120B-A12B.toml

@@ -10,7 +10,7 @@ type = "toggle"
 field = "reasoning_content"
 
 [cost]
-input = 0.06
+input = 0.3
 output = 0.75
 cache_read = 0.06
 

providers/baseten/models/zai-org/GLM-4.7.toml

@@ -17,7 +17,7 @@ type = "toggle"
 field = "reasoning_content"
 
 [cost]
-input = 0.12
+input = 0.6
 output = 2.2
 cache_read = 0.12
 

providers/baseten/models/zai-org/GLM-5.2.toml

@@ -0,0 +1,16 @@
+base_model = "zhipuai/glm-5.2"
+name = "GLM 5.2"
+
+[[reasoning_options]]
+type = "toggle"
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.5
+output = 4.5
+cache_read = 0.3
+
+[limit]
+context = 131_072

providers/claudinio/models/claudius.toml

@@ -0,0 +1,21 @@
+name = "Claudius"
+release_date = "2026-05-12"
+last_updated = "2026-05-12"
+attachment = true
+reasoning = true
+tool_call = true
+open_weights = false
+knowledge = "2026-05"
+
+[cost]
+input = 3.00
+output = 8.00
+cache_read = 0.900
+
+[limit]
+context = 256_000
+output = 64_000
+
+[modalities]
+input = ["text", "image", "audio", "video"]
+output = ["text"]

providers/cloudflare-workers-ai/models/@cf/zai-org/glm-5.2.toml

@@ -0,0 +1,11 @@
+base_model = "zhipuai/glm-5.2"
+name = "Glm 5.2"
+
+[cost]
+input = 1.4
+output = 4.4
+cache_read = 0.26
+
+[limit]
+context = 262_144
+output = 262_144

providers/cohere/models/command-r7b-12-2024.toml

@@ -1,7 +1,7 @@
 name = "Command R7B"
 family = "command-r"
-release_date = "2024-02-27"
-last_updated = "2024-02-27"
+release_date = "2024-12-02"
+last_updated = "2024-12-02"
 attachment = false
 reasoning = false
 temperature = true

providers/cortecs/models/devstral-small-2512.toml

@@ -1,22 +0,0 @@
-name = "Devstral Small 2 2512"
-release_date = "2025-12-09"
-last_updated = "2025-12-09"
-knowledge = "2025-12"
-attachment = false
-reasoning = false
-tool_call = true
-temperature = true
-open_weights = true
-
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/cortecs/models/glm-5-turbo.toml

@@ -0,0 +1,7 @@
+base_model = "zhipuai/glm-5-turbo"
+
+[cost]
+input = 1.235
+output = 4.118
+cache_read = 0.308
+cache_write = 1.544

providers/cortecs/models/glm-5.2.toml

@@ -0,0 +1,10 @@
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [{ type = "effort", values = ["high", "max"]}]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.44
+output = 4.53
+cache_read = 0.39

providers/cortecs/models/glm-5v-turbo.toml

@@ -0,0 +1,7 @@
+base_model = "zhipuai/glm-5v-turbo"
+
+[cost]
+input = 1.235
+output = 4.118
+cache_read = 0.308
+cache_write = 1.544

providers/cortecs/models/gpt-5.4.toml

@@ -1,5 +1,6 @@
 base_model = "openai/gpt-5.4"
 base_model_omit = ["limit.input"]
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 3

providers/cortecs/models/gpt-oss-120b.toml

@@ -5,6 +5,7 @@ last_updated = "2025-08-05"
 knowledge = "2024-01"
 attachment = false
 reasoning = false
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 tool_call = true
 temperature = true
 open_weights = true
@@ -20,4 +21,4 @@ output = 128_000
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/cortecs/models/hermes-4-70b.toml

@@ -4,6 +4,7 @@ last_updated = "2025-08-26"
 knowledge = "2023-12"
 attachment = false
 reasoning = true
+reasoning_options = []
 tool_call = true
 temperature = true
 open_weights = true

providers/cortecs/models/intellect-3.toml

@@ -4,6 +4,7 @@ last_updated = "2025-11-26"
 knowledge = "2025-11"
 attachment = true
 reasoning = true
+reasoning_options = []
 tool_call = true
 temperature = true
 open_weights = true
@@ -19,4 +20,4 @@ output = 128_000
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/cortecs/models/kimi-k2-thinking.toml

@@ -4,6 +4,7 @@ last_updated = "2025-12-08"
 knowledge = "2025-12"
 attachment = true
 reasoning = true
+reasoning_options = []
 tool_call = true
 temperature = true
 open_weights = true
@@ -21,4 +22,4 @@ output = 262_000
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

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

@@ -5,6 +5,7 @@ last_updated = "2026-01-27"
 knowledge = "2025-01"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true

providers/cortecs/models/kimi-k2.6.toml

@@ -4,6 +4,7 @@ release_date = "2026-04-17"
 last_updated = "2026-04-17"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true

providers/cortecs/models/kimi-k2.7-code.toml

@@ -0,0 +1,10 @@
+base_model = "moonshotai/kimi-k2.7-code"
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.28
+output = 4.63
+cache_read = 0.32

providers/cortecs/models/llama-3.3-70b-instruct.toml

@@ -2,6 +2,7 @@ base_model = "meta/llama-3.3-70b-instruct"
 name = "Llama 3.3 70B Instruct"
 attachment = false
 reasoning = true
+reasoning_options = []
 
 [cost]
 input = 0.089

providers/cortecs/models/llama-4-maverick.toml

@@ -0,0 +1,7 @@
+base_model = "meta/llama-4-maverick-17b-instruct"
+
+[cost]
+input = 0.124
+output = 0.603
+cache_read = 0.03
+cache_write = 0.151

providers/cortecs/models/minimax-m3.toml

@@ -0,0 +1,6 @@
+base_model = "minimax/MiniMax-M3"
+
+[cost]
+input = 0.355
+output = 1.775
+cache_read = 0.089

providers/cortecs/models/mixtral-8x7B-instruct-v0.1.toml

@@ -4,6 +4,7 @@ last_updated = "2023-12-11"
 knowledge = "2023-09"
 attachment = false
 reasoning = true
+reasoning_options = []
 tool_call = false
 temperature = true
 open_weights = true

providers/cortecs/models/nemotron-3-super-120b-a12b.toml

@@ -5,6 +5,7 @@ last_updated = "2026-03-11"
 knowledge = "2025-12"
 attachment = false
 reasoning = true
+reasoning_options = []
 tool_call = true
 temperature = true
 open_weights = true

providers/crof/models/deepseek-v4-pro-lightning.toml

@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro"
+name = "DeepSeek V4 Pro Lightning"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 
 [interleaved]

providers/crof/models/glm-5.2.toml

@@ -0,0 +1,16 @@
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.5
+output = 2.2
+cache_read = 0.08
+
+[limit]
+output = 131_072
+
+[provider]
+npm = "@ai-sdk/openai-compatible"

providers/crof/models/greg-2-super.toml

@@ -1,6 +1,6 @@
-name = "Greg 1 Super"
-release_date = "2026-01-27"
-last_updated = "2026-01-27"
+name = "Greg 2 Super"
+release_date = "2026-06-14"
+last_updated = "2026-06-14"
 attachment = false
 reasoning = false
 temperature = true
@@ -8,9 +8,9 @@ tool_call = false
 open_weights = false
 
 [cost]
-input = 1.00
-output = 5.00
-cache_read = 0.20
+input = 1.5
+output = 5
+cache_read = 0.25
 
 [limit]
 context = 229_376

providers/crof/models/greg-2-ultra.toml

@@ -1,6 +1,6 @@
-name = "Greg 1 Normal"
-release_date = "2026-01-27"
-last_updated = "2026-01-27"
+name = "Greg 2 Ultra"
+release_date = "2026-06-14"
+last_updated = "2026-06-14"
 attachment = false
 reasoning = false
 temperature = true
@@ -8,9 +8,9 @@ tool_call = false
 open_weights = false
 
 [cost]
-input = 0.10
-output = 0.30
-cache_read = 0.02
+input = 3
+output = 10
+cache_read = 0.5
 
 [limit]
 context = 229_376
@@ -18,4 +18,4 @@ output = 229_376
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/crof/models/kimi-k2.7-code.toml

@@ -0,0 +1,13 @@
+base_model = "moonshotai/kimi-k2.7-code"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.55
+output = 2.25
+cache_read = 0.05
+
+[provider]
+npm = "@ai-sdk/openai-compatible"

providers/firepass/logo.svg

@@ -1,3 +0,0 @@
-<svg width="24" height="24" viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg">
-<path d="M14.45 5.74999L11.9991 11.6956L9.54562 5.74999H7.97237L10.6604 12.2495C10.7678 12.5138 10.9515 12.7402 11.1882 12.8996C11.4248 13.059 11.7036 13.1442 11.9889 13.1444C12.2742 13.1446 12.5531 13.0597 12.79 12.9006C13.0268 12.7415 13.2109 12.5154 13.3186 12.2512L16.0232 5.74999H14.45ZM15.4965 14.808L19.98 10.2195L19.3684 8.75911L14.4719 13.7807C14.272 13.9856 14.1369 14.2448 14.0835 14.5261C14.0301 14.8073 14.0608 15.098 14.1718 15.3619C14.2807 15.624 14.4648 15.848 14.7008 16.0056C14.9369 16.1632 15.2144 16.2473 15.4983 16.2474L15.5 16.25L22.5 16.2325L21.8884 14.7721L15.4983 14.808H15.4965ZM4.02 10.216L4.63162 8.75561L9.52813 13.7772C9.93763 14.1964 10.0557 14.8176 9.82825 15.3584C9.71925 15.6204 9.53511 15.8443 9.29905 16.0019C9.06299 16.1595 8.78557 16.2437 8.50175 16.2439L1.50175 16.2281L1.5 16.2299L2.11163 14.7695L8.50175 14.8062L4.02 10.216Z" fill="currentColor"/>
-</svg>

providers/firepass/models/accounts/fireworks/routers/kimi-k2p6-turbo.toml

@@ -1,28 +0,0 @@
-name = "Kimi K2.6 Turbo"
-family = "kimi-thinking"
-release_date = "2026-04-17"
-last_updated = "2026-04-17"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = true
-
-[[reasoning_options]]
-type = "toggle"
-
-[cost]
-input = 0
-output = 0
-cache_read = 0
-
-[limit]
-context = 262_000
-output = 262_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[interleaved]
-field = "reasoning_content"

providers/firepass/provider.toml

@@ -1,5 +0,0 @@
-name = "Fireworks (Firepass)"
-env = ["FIREPASS_API_KEY"]
-npm = "@ai-sdk/openai-compatible"
-api = "https://api.fireworks.ai/inference/v1/"
-doc = "https://docs.fireworks.ai/firepass"

providers/fireworks-ai/models/accounts/fireworks/models/deepseek-v4-flash.toml

@@ -1,4 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash"
+last_updated = "2026-06-16"
 
 [[reasoning_options]]
 type = "toggle"
@@ -13,4 +14,4 @@ field = "reasoning_content"
 [cost]
 input = 0.14
 output = 0.28
-cache_read = 0.03
+cache_read = 0.028

providers/fireworks-ai/models/accounts/fireworks/models/glm-5p2.toml

@@ -0,0 +1,32 @@
+name = "GLM 5.2"
+family = "glm"
+release_date = "2026-06-16"
+last_updated = "2026-06-16"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 1.40
+output = 4.40
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[interleaved]
+field = "reasoning_content"

providers/fireworks-ai/models/accounts/fireworks/models/gpt-oss-120b.toml

@@ -1,7 +1,7 @@
 name = "GPT OSS 120B"
 family = "gpt-oss"
 release_date = "2025-08-05"
-last_updated = "2025-08-05"
+last_updated = "2026-06-16"
 attachment = false
 reasoning = true
 temperature = true

providers/fireworks-ai/models/accounts/fireworks/models/kimi-k2p7-code.toml

@@ -1,5 +1,12 @@
-base_model = "moonshotai/kimi-k2.7-code"
+name = "Kimi K2.7 Code"
+family = "kimi-k2"
+release_date = "2026-06-12"
+last_updated = "2026-06-16"
+attachment = true
+reasoning = true
 temperature = true
+tool_call = true
+open_weights = true
 
 [[reasoning_options]]
 type = "toggle"

providers/fireworks-ai/models/accounts/fireworks/routers/kimi-k2p7-code-fast.toml

@@ -1,13 +1,19 @@
-base_model = "moonshotai/kimi-k2.7-code"
 name = "Kimi K2.7 Code Fast"
+family = "kimi-k2"
+release_date = "2026-06-12"
+last_updated = "2026-06-16"
+attachment = true
+reasoning = true
 temperature = true
+tool_call = true
+open_weights = true
 
 [[reasoning_options]]
 type = "toggle"
 
 [cost]
 cache_read = 0.38
-input = 2.00
+input = 1.90
 output = 8.00
 
 [limit]

providers/frogbot/models/kimi-k2-6.toml

@@ -3,6 +3,7 @@ release_date = "1970-01-01"
 last_updated = "1970-01-01"
 attachment = false
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 structured_output = true

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

@@ -3,6 +3,7 @@ release_date = "1970-01-01"
 last_updated = "1970-01-01"
 attachment = false
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 structured_output = true

providers/frogbot/models/qwen-3-6-plus.toml

@@ -2,6 +2,7 @@ name = "Qwen 3.6 Plus"
 family = "qwen"
 attachment = true
 reasoning = true
+reasoning_options = []
 tool_call = true
 temperature = true
 release_date = "2026-04-02"

providers/github-copilot/models/claude-fable-5.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-fable-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 10

providers/github-copilot/models/claude-haiku-4.5.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-haiku-4-5"
-reasoning_options = [{ type = "budget_tokens", min = 1_024, max = 32_000 }]
 
 [cost]
 input = 1

providers/github-copilot/models/claude-opus-4.5.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-opus-4-5"
-reasoning_options = [{ type = "budget_tokens", min = 1_024, max = 32_000 }]
 
 [cost]
 input = 5

providers/github-copilot/models/claude-opus-4.6.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-opus-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }]
 
 [cost]
 input = 5

providers/github-copilot/models/claude-opus-4.7.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-opus-4-7"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 5

providers/github-copilot/models/claude-opus-4.8.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
 
 [cost]
 input = 5

providers/github-copilot/models/claude-sonnet-4.5.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-sonnet-4-5"
-reasoning_options = [{ type = "budget_tokens", min = 1_024, max = 32_000 }]
 
 [cost]
 input = 3

providers/github-copilot/models/claude-sonnet-4.6.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-sonnet-4-6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024, max = 32_000 }]
 
 [cost]
 input = 3

providers/github-copilot/models/claude-sonnet-4.toml

@@ -1,5 +1,4 @@
 base_model = "anthropic/claude-sonnet-4-0"
-reasoning_options = []
 
 [cost]
 input = 3

providers/github-copilot/models/gemini-2.5-pro.toml

@@ -1,4 +1,5 @@
 base_model = "google/gemini-2.5-pro"
+reasoning_options = [{ type = "budget_tokens", min = 128, max = 32_768 }]
 
 [cost]
 input = 1.25

providers/github-copilot/models/gemini-3-flash-preview.toml

@@ -1,4 +1,5 @@
 base_model = "google/gemini-3-flash-preview"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 256, max = 32_000 }]
 
 [cost]
 input = 0.5

providers/github-copilot/models/gemini-3.1-pro-preview.toml

@@ -1,4 +1,5 @@
 base_model = "google/gemini-3.1-pro-preview"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 256, max = 32_000 }]
 
 [cost]
 input = 2