Are team members even able to use different models?

[nfjdu]

I have Dev team member with setup like

## Model

- **Preferred:** claude-opus-4.6
- **Rationale:** Assigned per user request (Option A): Dev must use Claude Opus 4.6
- **Fallback:** gpt-5-mini

and its always using gpt-5-mini... So what im doing wrong, potentially?
What i did - just run squad init , told to setup a team like Lead(gpt-5-mini), Dev(claude-opus-4.6), QA(gpt-5-mini), DocsWriter(gpt-5-mini). But its not working like that.. All agents just use 'current model' selected in copilot cli session, or if im working from squad cli tool i have no idea how it selects model that will be used, either automatically or always uses fallback model

casting/registry.json:

{
  "agents": {
    "neo": {
      "created_at": "2026-03-15T09:53:50.627Z",
      "persistent_name": "Neo",
      "universe": "The Matrix",
      "status": "active"
    },
    "trinity": {
      "created_at": "2026-03-15T09:53:50.627Z",
      "persistent_name": "Trinity",
      "universe": "The Matrix",
      "status": "active",
      "model": "claude-opus-4.6"
    },
    "morpheus": {
      "created_at": "2026-03-15T09:53:50.627Z",
      "persistent_name": "Morpheus",
      "universe": "The Matrix",
      "status": "active"
    },
    "oracle": {
      "created_at": "2026-03-15T09:53:50.627Z",
      "persistent_name": "Oracle",
      "universe": "The Matrix",
      "status": "active"
    },
    "scribe": {
      "created_at": "2026-03-15T09:53:50.627Z",
      "persistent_name": "Scribe",
      "universe": "The Matrix",
      "status": "active"
    },
    "ralph": {
      "created_at": "2026-03-15T09:53:50.627Z",
      "persistent_name": "Ralph",
      "universe": "The Matrix",
      "status": "active"
    }
  }
}

[bradygaster]

This is a real limitation we've been working on. The ## Model section in charters expresses a preference, but enforcement depends on how the Copilot runtime handles model selection.

A few things to check:

1. Model availability — Make sure claude-opus-4.6 is available in your Copilot subscription. If it's not, the runtime falls back to whatever is available (which may be gpt-5-mini).
2. Agent spawn context — When Squad's coordinator spawns an agent via the task tool, it passes the model preference. But the runtime may override it based on availability and quotas.

We just merged PR #491 which adds Layer 0 persistent model config via .squad/config.json, giving you another way to set model preferences at the team and per-agent level. Give that a try when v0.9.0 drops this weekend.

That said, the honest answer is that model selection is ultimately up to the Copilot runtime, and Squad can only request a specific model. We're tracking this as a known pain point.

[nfjdu]

updated package, tried to setup team again from scratch on a new project. still was no able to specify some agent to use some specific model.. :( im sure that model i available for me because im using it from copilot cli. i have copilot pro plan. so premium requests are limited, but models e.g. opus are available. So.. will be waiting new releases seems like 😅