Repository navigation
Question: Set model v.s Squad model in Copilot? #494
Description
Does it matter if I set the model (/models) in Copilot CLI? Will that affect what model is used underneeth? It would be great if that would control sometheing or at least if we would get some outpout saying "your chosen model will not have any affect since its not set in Squad".. or just doucumented so we can be sure what is to be expected.
And regarding the medium, high, extra hight options we get in the Copilot CLI.. are they used in Sqad at all (if e.g. 5.4 is used)?
Activity
jchannon commentedon Mar 25, 2026
What would be amazing would be in the
defineAgentstuff we could specify the models. For example, have one agent use a cheap model to plan, another model to execute and another model to review.Is this possible or on the 'roadmap'?
docs(ai-team): Wave D batch 1 implementation complete
bradygaster commentedon Mar 26, 2026
Great question, @sturlath! Squad's model selection uses a 4-layer resolution chain: config.json → session directive → charter preference → task-aware auto-selection. The Copilot CLI's model picker IS used (as a session directive), but Squad's persistent config and per-agent overrides take priority. With PR #628, we're adding \squad config model\ for explicit CLI control, giving you full transparency over selection. The Copilot CLI's 'medium/high/extra high' tiers control underlying capacity, not Squad's routing. Thanks for surfacing this! 🎯