Expose thinking capability in Models API

[milindmore22]

Which version of LM Studio?
Example: LM Studio Version 0.4.6+1 (0.4.6+1)

Which operating system?
Mac OS Tahoe 26.3.1

What is the bug?
Expose reasoning/thinking capabilities for the models that provides it, it can be helpful in determining to turn it off for specific models

eg:
GET http://127.0.0.1:1234/api/v1/models

Current Response

"capabilities": {
        "vision": true,
        "trained_for_tool_use": true
      },

Updated response - This will help in determining where we can turn off reasoning

"capabilities": {
        "vision": true,
        "trained_for_tool_use": true,
       "thinking": true
      },

Screenshots
Not Applicable

Logs
No Logs adding my current models

{
  "models": [
    {
      "type": "llm",
      "publisher": "google",
      "key": "google/gemma-3-4b",
      "display_name": "Gemma 3 4B",
      "architecture": "gemma3",
      "quantization": {
        "name": "4bit",
        "bits_per_weight": 4
      },
      "size_bytes": 3034755901,
      "params_string": "4B",
      "loaded_instances": [],
      "max_context_length": 131072,
      "format": "mlx",
      "capabilities": {
        "vision": true,
        "trained_for_tool_use": false
      },
      "description": null,
      "variants": [
        "google/gemma-3-4b@4bit"
      ],
      "selected_variant": "google/gemma-3-4b@4bit"
    },
    {
      "type": "llm",
      "publisher": "qwen",
      "key": "qwen/qwen3.5-9b",
      "display_name": "Qwen3.5 9B",
      "architecture": "qwen35",
      "quantization": {
        "name": "Q4_K_M",
        "bits_per_weight": 4
      },
      "size_bytes": 6548927017,
      "params_string": "9B",
      "loaded_instances": [
        {
          "id": "qwen/qwen3.5-9b",
          "config": {
            "context_length": 4096,
            "eval_batch_size": 512,
            "flash_attention": true,
            "offload_kv_cache_to_gpu": true
          }
        }
      ],
      "max_context_length": 262144,
      "format": "gguf",
      "capabilities": {
        "vision": true,
        "trained_for_tool_use": true
      },
      "description": null,
      "variants": [
        "qwen/qwen3.5-9b@q4_k_m"
      ],
      "selected_variant": "qwen/qwen3.5-9b@q4_k_m"
    },
    {
      "type": "embedding",
      "publisher": "nomic-ai",
      "key": "text-embedding-nomic-embed-text-v1.5",
      "display_name": "Nomic Embed Text v1.5",
      "quantization": {
        "name": "Q4_K_M",
        "bits_per_weight": 4
      },
      "size_bytes": 84106624,
      "params_string": null,
      "loaded_instances": [],
      "max_context_length": 2048,
      "format": "gguf"
    }
  ]
}

To Reproduce
Steps to reproduce the behavior:

1. Start LM Studio
2. Install models such as qwen3.5 9b
3. Start Local API server
4. Visit http://127.0.0.1:1234/api/v1/models

[milindmore22]

Closing this as it has been added as feature in 0.4.8 release.
Now the reasoning is exposed as follow

"capabilities": {
        "vision": false,
        "trained_for_tool_use": true,
        "reasoning": {
          "allowed_options": [
            "off",
            "on"
          ],
          "default": "on"
        }
      }

complete response

{
  "models": [
    {
      "type": "llm",
      "publisher": "Jackrong",
      "key": "qwen3.5-27b-claude-4.6-opus-reasoning-distilled",
      "display_name": "Qwen3.5 27B",
      "architecture": "qwen35",
      "quantization": {
        "name": "Q2_K",
        "bits_per_weight": 2
      },
      "size_bytes": 11052986271,
      "params_string": "27B",
      "loaded_instances": [],
      "max_context_length": 262144,
      "format": "gguf",
      "capabilities": {
        "vision": true,
        "trained_for_tool_use": true
      },
      "description": null
    },
    {
      "type": "llm",
      "publisher": "nvidia",
      "key": "nvidia/nemotron-3-nano-4b",
      "display_name": "Nemotron 3 Nano 4B",
      "architecture": "nemotron_h",
      "quantization": {
        "name": "Q4_K_M",
        "bits_per_weight": 4
      },
      "size_bytes": 2837173509,
      "params_string": "4.0B",
      "loaded_instances": [],
      "max_context_length": 1048576,
      "format": "gguf",
      "capabilities": {
        "vision": false,
        "trained_for_tool_use": true,
        "reasoning": {
          "allowed_options": [
            "off",
            "on"
          ],
          "default": "on"
        }
      },
      "description": null,
      "variants": [
        "nvidia/nemotron-3-nano-4b@q4_k_m"
      ],
      "selected_variant": "nvidia/nemotron-3-nano-4b@q4_k_m"
    },
    {
      "type": "llm",
      "publisher": "mistralai",
      "key": "mistralai/ministral-3-3b",
      "display_name": "Ministral 3 3B",
      "architecture": "mistral3",
      "quantization": {
        "name": "Q4_K_M",
        "bits_per_weight": 4
      },
      "size_bytes": 2986817071,
      "params_string": "3B",
      "loaded_instances": [],
      "max_context_length": 262144,
      "format": "gguf",
      "capabilities": {
        "vision": true,
        "trained_for_tool_use": true
      },
      "description": null,
      "variants": [
        "mistralai/ministral-3-3b@q4_k_m"
      ],
      "selected_variant": "mistralai/ministral-3-3b@q4_k_m"
    },
    {
      "type": "llm",
      "publisher": "google",
      "key": "google/gemma-3-4b",
      "display_name": "Gemma 3 4B",
      "architecture": "gemma3",
      "quantization": {
        "name": "4bit",
        "bits_per_weight": 4
      },
      "size_bytes": 3034755901,
      "params_string": "4B",
      "loaded_instances": [],
      "max_context_length": 131072,
      "format": "mlx",
      "capabilities": {
        "vision": true,
        "trained_for_tool_use": false
      },
      "description": null,
      "variants": [
        "google/gemma-3-4b@4bit"
      ],
      "selected_variant": "google/gemma-3-4b@4bit"
    },
    {
      "type": "embedding",
      "publisher": "nomic-ai",
      "key": "text-embedding-nomic-embed-text-v1.5",
      "display_name": "Nomic Embed Text v1.5",
      "quantization": {
        "name": "Q4_K_M",
        "bits_per_weight": 4
      },
      "size_bytes": 84106624,
      "params_string": null,
      "loaded_instances": [],
      "max_context_length": 2048,
      "format": "gguf"
    }
  ]
}

Docs are yet to be updated: https://lmstudio.ai/docs/developer/rest/list