Skip to content

Switching through agents doesn't change model #2770

Closed

Description

@lucaorio

It appears that tabbing through the agents or selecting them from the related dialog does not update the models.

Sometimes, the model description is stuck on Grok on the first prompt screen, but I don't have Grok defined anywhere in my config. Then, I prompt:

What LLM model are you?

It answers:

I'm from Anthropic [...]

I'm very confused; below is my opencode.json for a fresh, brand-new installation of v0.11.3. Am I getting anything wrong here?

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "lmstudio": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LM Studio",
      "options": {
        "baseURL": "http://127.0.0.1:1234"
      },
      "models": {
        "openai/gpt-oss-20b": {
          "name": "GPT-OSS-20b",
          "limit": {
            "context": 8000,
            "output": 2048
          }
        },
        "qwen3-coder-30b-a3b-instruct-mlx": {
          "name": "Qwen3-Coder-30b",
          "limit": {
            "context": 8000,
            "output": 2048
          }
        },
      }
    }
  },
  "agent": {
    "build": {
      "mode": "primary",
      "model": "qwen3-coder-30b-a3b-instruct-mlx",
      "tools": {
        "write": true,
        "edit": true,
        "bash": true
      }
    },
    "plan": {
      "mode": "primary",
      "model": "openai/gpt-oss-20b",
      "tools": {
        "write": false,
        "edit": false,
        "bash": false
      }
    },
  }
}

Activity

  1. lucaorio commented on Sep 25, 2025

    @lucaorio
    Author

    Closing this as I figured out what was wrong. For anyone who stumbles upon this: In the agent definition, models' names must include the provider (i.e., "model": "lmstudio/qwen3-coder-30b-a3b-instruct-mlx").

  2. arsham commented on Sep 25, 2025

    @arsham

    I would like to add another tip here. When you want to see what is the model's handle, in a fresh chat choose that particular model and say "hi". Under the response you will see the handle something like "hf:zai-org/GLM-4.5" or "claude-opus-4-1-20250805" or "gemini-2.5-pro". Then the model you need to set must always follow this pattern: /.

    Here are a few examples:

    model: google/gemini-1.5-pro
    model: anthropic/claude-3-5-haiku-20241022
    model: anthropic/claude-opus-4-1-20250805
    model: synthetic/hf:deepseek-ai/DeepSeek-R1-0528
    model: synthetic/hf:zai-org/GLM-4.5
    model: synthetic/hf:Qwen/Qwen3-235B-A22B-Thinking-2507
    
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions