Cannot select reasoning_effort=high; model claude-opus-4.7-high rejects requests with 400 error

[dezgit2025]

Summary

When the model is set to claude-opus-4.7-high, every request fails with a 400 error from CAPI because the CLI sends reasoning_effort: "medium", but that model only supports high. There is no UI/setting to change reasoning effort, so the model is effectively unusable.

Repro

1. Launch copilot (v1.0.34)
2. Switch model to claude-opus-4.7-high
3. Send any prompt (e.g. ping)

Actual

● Request failed (transient_bad_request). Retrying...
● Request failed (transient_bad_request). Retrying...
✗ Execution failed: CAPIError: 400 reasoning_effort "medium" is not supported by model claude-opus-4.7-high; supported values: [high] (Request ID: E19A:1909B6:23466F8:28A69BC:69F56D5A)
● Model changed to: claude-opus-4.7-1m-internal

The CLI auto-falls-back to a different model.

Expected

Either:

• The CLI should automatically send reasoning_effort: "high" when the selected model only supports that value, or
• Expose a setting to choose reasoning effort, or
• Hide/disable models in the picker that would always 400 with default settings.

Environment

• GitHub Copilot CLI 1.0.34
• macOS (Darwin)
• Model attempted: claude-opus-4.7-high

Notes

No visible selection in the model picker / settings to change reasoning effort, so there is no user workaround other than picking a different model.

[dezgit2025]

Can confirm local repro on Copilot CLI 1.0.34. Help text shows --reasoning-effort and --effort options accepting high/xhigh, but requests still transmit reasoning_effort: "medium", causing 400 errors for high/xhigh model variants like claude-opus-4.7-high.