Audit Together AI models and reasoning controls

packages/core/test/togetherai-generated.test.ts

@@ -0,0 +1,101 @@
+import { expect, test } from "bun:test";
+import path from "node:path";
+
+import { generate } from "../src/index.js";
+
+const expected = {
+  "LiquidAI/LFM2-24B-A2B": ["active", false, undefined],
+  "MiniMaxAI/MiniMax-M2.5": ["deprecated", true, undefined],
+  "MiniMaxAI/MiniMax-M2.7": ["active", true, []],
+  "Qwen/Qwen2.5-7B-Instruct-Turbo": ["active", false, undefined],
+  "Qwen/Qwen3-235B-A22B-Instruct-2507-tput": ["active", false, undefined],
+  "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": ["deprecated", false, undefined],
+  "Qwen/Qwen3-Coder-Next-FP8": ["deprecated", false, undefined],
+  "Qwen/Qwen3.5-397B-A17B": ["active", true, [{ type: "toggle" }]],
+  "Qwen/Qwen3.5-9B": ["active", true, [{ type: "toggle" }]],
+  "Qwen/Qwen3.6-Plus": ["active", true, [{ type: "toggle" }]],
+  "Qwen/Qwen3.7-Max": ["active", false, undefined],
+  "deepcogito/cogito-v2-1-671b": ["active", true, [{ type: "toggle" }]],
+  "deepseek-ai/DeepSeek-R1": ["deprecated", true, []],
+  "deepseek-ai/DeepSeek-V3": ["deprecated", false, undefined],
+  "deepseek-ai/DeepSeek-V3-1": ["deprecated", true, [{ type: "toggle" }]],
+  "deepseek-ai/DeepSeek-V4-Pro": [
+    "active",
+    true,
+    [{ type: "toggle" }, { type: "effort", values: ["high", "max"] }],
+  ],
+  "essentialai/Rnj-1-Instruct": ["active", false, undefined],
+  "google/gemma-3n-E4B-it": ["active", false, undefined],
+  "google/gemma-4-31B-it": ["active", true, undefined],
+  "meta-llama/Llama-3.3-70B-Instruct-Turbo": ["active", false, undefined],
+  "meta-llama/Meta-Llama-3-8B-Instruct-Lite": ["active", false, undefined],
+  "moonshotai/Kimi-K2.5": ["deprecated", true, [{ type: "toggle" }]],
+  "moonshotai/Kimi-K2.6": ["active", true, [{ type: "toggle" }]],
+  "nvidia/nemotron-3-ultra-550b-a55b": ["active", true, [{ type: "toggle" }]],
+  "openai/gpt-oss-120b": [
+    "active",
+    true,
+    [{ type: "effort", values: ["low", "medium", "high"] }],
+  ],
+  "openai/gpt-oss-20b": [
+    "active",
+    true,
+    [{ type: "effort", values: ["low", "medium", "high"] }],
+  ],
+  "pearl-ai/gemma-4-31b-it": ["active", true, undefined],
+  "zai-org/GLM-5": ["active", true, [{ type: "toggle" }]],
+  "zai-org/GLM-5.1": ["active", true, [{ type: "toggle" }]],
+} as const;
+
+test("generated Together matrix is exhaustive and resolved", async () => {
+  const root = path.join(import.meta.dirname, "..", "..", "..");
+  const providers = await generate(path.join(root, "providers"));
+  const models = providers.togetherai?.models;
+
+  expect(models).toBeDefined();
+  expect(Object.keys(models ?? {}).sort()).toEqual(Object.keys(expected).sort());
+
+  const active: string[] = [];
+  const deprecated: string[] = [];
+  const unresolved: string[] = [];
+  const fixed: string[] = [];
+
+  for (const [id, [lifecycle, reasoning, reasoningOptions]] of Object.entries(expected)) {
+    const model = models?.[id];
+    expect(model, id).toBeDefined();
+    expect(model?.status ?? "active", `${id} lifecycle`).toBe(lifecycle);
+    expect(model?.reasoning, `${id} reasoning`).toBe(reasoning);
+    expect(model?.reasoning_options, `${id} controls`).toEqual(reasoningOptions);
+
+    const encoded = JSON.stringify(model);
+    expect(encoded, `${id} resolved metadata`).not.toContain("base_model");
+
+    (lifecycle === "deprecated" ? deprecated : active).push(id);
+    if (reasoning && reasoningOptions === undefined) unresolved.push(id);
+    if (reasoningOptions?.length === 0) fixed.push(id);
+    if (!reasoning) expect(model?.reasoning_options, `${id} non-reasoning controls`).toBeUndefined();
+  }
+
+  expect(active).toHaveLength(22);
+  expect(deprecated).toHaveLength(7);
+  expect(unresolved.sort()).toEqual([
+    "MiniMaxAI/MiniMax-M2.5",
+    "google/gemma-4-31B-it",
+    "pearl-ai/gemma-4-31b-it",
+  ]);
+  expect(fixed.sort()).toEqual([
+    "MiniMaxAI/MiniMax-M2.7",
+    "deepseek-ai/DeepSeek-R1",
+  ]);
+
+  expect(models?.["deepseek-ai/DeepSeek-V4-Pro"]?.cost).toMatchObject({
+    input: 1.74,
+    output: 3.48,
+    cache_read: 0.2,
+  });
+  expect(models?.["google/gemma-4-31B-it"]).toMatchObject({
+    reasoning: true,
+    structured_output: true,
+    cost: { input: 0.39, output: 0.97 },
+  });
+});

providers/togetherai/models/LiquidAI/LFM2-24B-A2B.toml

@@ -0,0 +1,21 @@
+name = "LFM2-24B-A2B"
+family = "liquid"
+release_date = "2026-02-25"
+last_updated = "2026-02-25"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+open_weights = true
+
+[cost]
+input = 0.03
+output = 0.12
+
+[limit]
+context = 32_768
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/togetherai/models/MiniMaxAI/MiniMax-M2.5.toml

@@ -4,6 +4,7 @@ release_date = "2026-02-12"
 last_updated = "2026-02-12"
 attachment = false
 reasoning = true
+status = "deprecated"
 temperature = true
 tool_call = true
 open_weights = true

providers/togetherai/models/MiniMaxAI/MiniMax-M2.7.toml

@@ -4,6 +4,7 @@ release_date = "2026-03-18"
 last_updated = "2026-03-18"
 attachment = false
 reasoning = true
+reasoning_options = []
 temperature = true
 tool_call = true
 structured_output = true

providers/togetherai/models/Qwen/Qwen2.5-7B-Instruct-Turbo.toml

@@ -0,0 +1,22 @@
+name = "Qwen 2.5 7B Instruct Turbo"
+family = "qwen"
+release_date = "2024-09-19"
+last_updated = "2024-09-19"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.30
+output = 0.30
+
+[limit]
+context = 32_768
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/togetherai/models/Qwen/Qwen3-235B-A22B-Instruct-2507-tput.toml

@@ -3,7 +3,7 @@ family = "qwen"
 release_date = "2025-07-25"
 last_updated = "2025-07-25"
 attachment = false
-reasoning = true
+reasoning = false
 temperature = true
 knowledge = "2025-07"
 tool_call = true
@@ -19,4 +19,4 @@ output = 262_144
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/togetherai/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8.toml

@@ -4,6 +4,7 @@ release_date = "2025-07-23"
 last_updated = "2025-07-23"
 attachment = false
 reasoning = false
+status = "deprecated"
 temperature = true
 knowledge = "2025-04"
 tool_call = true

providers/togetherai/models/Qwen/Qwen3-Coder-Next-FP8.toml

@@ -3,7 +3,8 @@ family = "qwen"
 release_date = "2026-02-03"
 last_updated = "2026-02-03"
 attachment = false
-reasoning = true
+reasoning = false
+status = "deprecated"
 temperature = true
 knowledge = "2026-02-03"
 tool_call = true

providers/togetherai/models/Qwen/Qwen3.5-397B-A17B.toml

@@ -4,6 +4,7 @@ release_date = "2026-02-16"
 last_updated = "2026-02-16"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true

providers/togetherai/models/Qwen/Qwen3.5-9B.toml

@@ -0,0 +1,23 @@
+name = "Qwen3.5 9B"
+family = "qwen"
+release_date = "2026-03-03"
+last_updated = "2026-03-03"
+attachment = true
+reasoning = true
+reasoning_options = [{ type = "toggle" }]
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.17
+output = 0.25
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/togetherai/models/Qwen/Qwen3.6-Plus.toml

@@ -4,6 +4,7 @@ release_date = "2026-04-30"
 last_updated = "2026-04-30"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true

providers/togetherai/models/Qwen/Qwen3.7-Max.toml

@@ -3,7 +3,7 @@ family = "qwen"
 release_date = "2026-05-21"
 last_updated = "2026-05-21"
 attachment = false
-reasoning = true
+reasoning = false
 temperature = true
 tool_call = true
 open_weights = false

providers/togetherai/models/deepcogito/cogito-v2-1-671b.toml

@@ -0,0 +1,22 @@
+name = "Cogito v2.1 671B"
+family = "cogito"
+release_date = "2025-11-13"
+last_updated = "2025-11-13"
+attachment = false
+reasoning = true
+reasoning_options = [{ type = "toggle" }]
+temperature = true
+tool_call = false
+open_weights = false
+
+[cost]
+input = 1.25
+output = 1.25
+
+[limit]
+context = 163_840
+output = 163_840
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/togetherai/models/deepseek-ai/DeepSeek-R1.toml

@@ -4,6 +4,8 @@ release_date = "2025-01-20"
 last_updated = "2025-03-24"
 attachment = false
 reasoning = true
+reasoning_options = []
+status = "deprecated"
 temperature = true
 knowledge = "2024-07"
 tool_call = false

providers/togetherai/models/deepseek-ai/DeepSeek-V3-1.toml

@@ -4,6 +4,8 @@ release_date = "2025-08-21"
 last_updated = "2025-08-21"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
+status = "deprecated"
 temperature = true
 tool_call = true
 knowledge = "2025-08"

providers/togetherai/models/deepseek-ai/DeepSeek-V3.toml

@@ -3,7 +3,8 @@ family = "deepseek"
 release_date = "2024-12-26"
 last_updated = "2025-05-29"
 attachment = false
-reasoning = true
+reasoning = false
+status = "deprecated"
 temperature = true
 knowledge = "2024-07"
 tool_call = true

providers/togetherai/models/deepseek-ai/DeepSeek-V4-Pro.toml

@@ -4,6 +4,7 @@ release_date = "2026-04-24"
 last_updated = "2026-04-24"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
 structured_output = true
 temperature = true
 tool_call = true
@@ -13,8 +14,8 @@ open_weights = true
 field = "reasoning_content"
 
 [cost]
-input = 2.10
-output = 4.40
+input = 1.74
+output = 3.48
 cache_read = 0.20
 
 [limit]

providers/togetherai/models/google/gemma-3n-E4B-it.toml

@@ -0,0 +1,22 @@
+name = "Gemma 3N E4B Instruct"
+family = "gemma"
+release_date = "2025-05-20"
+last_updated = "2025-05-20"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.06
+output = 0.12
+
+[limit]
+context = 32_768
+output = 32_768
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/togetherai/models/google/gemma-4-31B-it.toml

@@ -4,14 +4,15 @@ release_date = "2026-04-07"
 last_updated = "2026-04-07"
 attachment = true
 reasoning = true
+structured_output = true
 temperature = true
 knowledge = "2025-01"
 tool_call = true
 open_weights = true
 
 [cost]
-input = 0.20
-output = 0.50
+input = 0.39
+output = 0.97
 
 [limit]
 context = 262_144

providers/togetherai/models/meta-llama/Meta-Llama-3-8B-Instruct-Lite.toml

@@ -0,0 +1,21 @@
+name = "Meta Llama 3 8B Instruct Lite"
+family = "llama"
+release_date = "2024-04-18"
+last_updated = "2024-04-18"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+open_weights = true
+
+[cost]
+input = 0.14
+output = 0.14
+
+[limit]
+context = 8_192
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/togetherai/models/moonshotai/Kimi-K2.5.toml

@@ -4,6 +4,8 @@ release_date = "2026-01-27"
 last_updated = "2026-01-27"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }]
+status = "deprecated"
 temperature = true
 tool_call = true
 knowledge = "2026-01"

providers/togetherai/models/moonshotai/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" }]
 structured_output = true
 temperature = true
 tool_call = true