chore(manifests): add outputModalities to model CLIs

Add outputModalities field to all model manifests and kimi-cli
to specify supported output media types for AI models.

Co-Authored-By: Claude <noreply@anthropic.com>

manifests/clis/kimi-cli.json

@@ -78,6 +78,12 @@
       "launchCommand": "kimi",
       "installPath": null
     },
+    {
+      "os": "Windows",
+      "installCommand": "uv tool install --python 3.13 kimi-cli",
+      "launchCommand": "kimi",
+      "installPath": null
+    },
     {
       "os": "Linux",
       "installCommand": "uv tool install --python 3.13 kimi-cli",

manifests/models/claude-haiku-4-5.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-02",
   "inputModalities": ["text", "image", "pdf"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/claude-opus-4-1.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": null,
   "inputModalities": ["text", "image", "pdf"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "reasoning", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": 74.4,

manifests/models/claude-opus-4-5.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-05",
   "inputModalities": ["text", "image", "pdf"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/claude-opus-4.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": null,
   "inputModalities": ["text", "image", "pdf"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "reasoning", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": 67.6,

manifests/models/claude-sonnet-4-5.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-01",
   "inputModalities": ["text", "image", "pdf"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "reasoning", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": 70.6,

manifests/models/claude-sonnet-4.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": null,
   "inputModalities": ["text", "image", "pdf"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": 64.93,

manifests/models/composer.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": null,
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": null,

manifests/models/deepseek-3-2.json

@@ -0,0 +1,73 @@
+{
+  "$schema": "../$schemas/model.schema.json",
+  "id": "deepseek-3-2",
+  "name": "DeepSeek V3.2",
+  "description": "DeepSeek V3.2 is a reasoning-first model with 128K context window, designed for agent tasks with integrated thinking in tool-use capabilities.",
+  "translations": {
+    "zh-Hans": {
+      "description": "DeepSeek V3.2 是推理优先的模型，具备 128K 上下文窗口，专为智能体任务设计，将思考直接集成到工具使用中。"
+    },
+    "de": {
+      "description": "DeepSeek V3.2 ist ein reasoning-first Modell mit 128K-Kontextfenster, optimiert fur Agent-Tasks mit integriertem Denken in Tool-Use-Fahigkeiten."
+    },
+    "ko": {
+      "description": "DeepSeek V3.2은 128K 컨텍스트 윈도를 갖춘 추론 우선 모델로, 도구 사용에 통합된 생각 기능을 갖춘 에이전트 작업을 위해 설계되었습니다."
+    },
+    "es": {
+      "description": "DeepSeek V3.2 es un modelo de razonamiento prioritario con ventana de contexto de 128K, disenado para tareas de agentes con capacidades de pensamiento integrado en herramienta."
+    },
+    "fr": {
+      "description": "DeepSeek V3.2 est un modele raisonnement-prioritaire avec une fenetre de contexte de 128K, concu pour les taches d'agents avec des capacites de reflechion integrees dans l'utilisation d'outils."
+    },
+    "id": {
+      "description": "DeepSeek V3.2 adalah model penalaran prioritas dengan jendela konteks 128K, dirancang untuk tugas agen dengan kemampuan berpikir terintegrasi dalam penggunaan alat."
+    },
+    "ja": {
+      "description": "DeepSeek V3.2は128Kコンテキストウィンドウを持つ推論優先モデルで、ツール使用に統合された思考機能を持つエージェントタスクのために設計されています。"
+    },
+    "pt": {
+      "description": "DeepSeek V3.2 e um modelo de raciocinio prioritario com janela de contexto de 128K, projetado para tarefas de agentes com capacidades de pensamento integradas em uso de ferramentas."
+    },
+    "ru": {
+      "description": "DeepSeek V3.2 - это модель с приоритетом рассуждения с окном контекста 128K, разработанная для задач агентов с интегрированным мышлением в использовании инструментов."
+    },
+    "tr": {
+      "description": "DeepSeek V3.2, 128K baglam penceresine sahip oncelikli muhakeme modelidir, arac kullaniminda entegre dusunme yeteneklerine sahip ortam gorevleri icin tasarlanmistir."
+    },
+    "zh-Hant": {
+      "description": "DeepSeek V3.2 是推理優先的模型，具備 128K 上下文視窗，專為智能體任務設計，將思考直接整合到工具使用中。"
+    }
+  },
+  "verified": false,
+  "websiteUrl": "https://www.deepseek.com",
+  "docsUrl": "https://api-docs.deepseek.com/quick_start/pricing",
+  "vendor": "DeepSeek",
+  "size": "671B",
+  "contextWindow": 128000,
+  "maxOutput": 8000,
+  "tokenPricing": {
+    "input": 0.28,
+    "output": 0.42,
+    "cache": 0.028
+  },
+  "releaseDate": "2025-12-01",
+  "lifecycle": "latest",
+  "knowledgeCutoff": null,
+  "inputModalities": ["text"],
+  "outputModalities": ["text"],
+  "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
+  "benchmarks": {
+    "sweBench": null,
+    "terminalBench": null,
+    "sciCode": null,
+    "liveCodeBench": null,
+    "mmmu": null,
+    "mmmuPro": null,
+    "webDevArena": null
+  },
+  "platformUrls": {
+    "huggingface": "https://huggingface.co/deepseek-ai/DeepSeek-V3.2",
+    "artificialAnalysis": null,
+    "openrouter": null
+  }
+}

manifests/models/deepseek-r1.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": null,
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/deepseek-v3-terminus.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": null,
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/gemini-2-5-flash.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2025-03",
   "inputModalities": ["text", "image", "pdf", "audio", "video"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": 28.73,

manifests/models/gemini-2-5-pro.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2025-05",
   "inputModalities": ["text", "image", "pdf", "audio", "video"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": 53.6,

manifests/models/gemini-3-flash.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-01",
   "inputModalities": ["text", "image", "pdf", "audio", "video"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/gemini-3-pro.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-01",
   "inputModalities": ["text", "image", "pdf", "audio", "video"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": 74.2,

manifests/models/glm-4-6.json

@@ -54,13 +54,14 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": null,
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": 55.4,
     "terminalBench": 0.245,
     "sciCode": null,
     "liveCodeBench": null,
-    "mmmu": 68.0,
+    "mmmu": 68,
     "mmmuPro": null,
     "webDevArena": null
   },

manifests/models/glm-4-6v.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-12",
   "inputModalities": ["text", "image", "video"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/glm-4-7.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-12",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/gpt-4-1.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2024-06-01",
   "inputModalities": ["text", "image"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": 39.58,

manifests/models/gpt-4o.json

@@ -54,6 +54,7 @@
   "lifecycle": "deprecated",
   "knowledgeCutoff": "2023-10-01",
   "inputModalities": ["text", "image"],
+  "outputModalities": ["text"],
   "capabilities": ["tool-choice", "structured-outputs", "function-calling"],
   "benchmarks": {
     "sweBench": 21.62,

manifests/models/gpt-5-1-codex.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2024-09-30",
   "inputModalities": ["text", "image"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/gpt-5-1.json

@@ -54,13 +54,14 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2024-09-30",
   "inputModalities": ["text", "image"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
-    "sweBench": 66.0,
+    "sweBench": 66,
     "terminalBench": 0.476,
     "sciCode": null,
     "liveCodeBench": null,
-    "mmmu": 76.0,
+    "mmmu": 76,
     "mmmuPro": null,
     "webDevArena": null
   },

manifests/models/gpt-5-2.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-08-31",
   "inputModalities": ["text", "image"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/gpt-5-codex.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2024-09-30",
   "inputModalities": ["text", "image"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/gpt-5.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2024-09-30",
   "inputModalities": ["text", "image"],
+  "outputModalities": ["text"],
   "capabilities": ["tool-choice", "structured-outputs", "function-calling", "reasoning"],
   "benchmarks": {
     "sweBench": 71.8,
@@ -62,7 +63,7 @@
     "liveCodeBench": 28.7,
     "mmmu": 74.4,
     "mmmuPro": null,
-    "webDevArena": 148.0
+    "webDevArena": 148
   },
   "platformUrls": {
     "huggingface": null,

manifests/models/grok-code-fast-1.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": null,
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/kat-coder-pro-v1.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-11",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/kimi-k2-0905.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-09",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": 43.8,

manifests/models/kimi-k2-thinking.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-11",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/minimax-m2-1.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-12",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": null,

manifests/models/minimax-m2.json

@@ -54,6 +54,7 @@
   "lifecycle": "maintained",
   "knowledgeCutoff": "2025-10",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs", "reasoning"],
   "benchmarks": {
     "sweBench": 61,

manifests/models/qwen3-coder-30b-a3b.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-07",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": null,

manifests/models/qwen3-coder-480b-a35b.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-07",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": null,

manifests/models/qwen3-coder-plus.json

@@ -54,6 +54,7 @@
   "lifecycle": "latest",
   "knowledgeCutoff": "2025-07",
   "inputModalities": ["text"],
+  "outputModalities": ["text"],
   "capabilities": ["function-calling", "tool-choice", "structured-outputs"],
   "benchmarks": {
     "sweBench": null,