You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Misleading model_not_supported error after content-filter response #5309
Copilot CLI reports Execution failed: 400 The requested model is not supported even though the same session successfully makes requests with Claude Opus 5.5. I also reproduced the error with additional models, including Auto. Resetting subagent preferences and restarting with debug logging did not resolve it.
In the debug session, the Opus 5.5 request immediately before the failure was recorded with finishReason: content_filter . The subsequent request returned model_not_supported ( invalid_request_error , param model ). This suggests the 400 may be a misleading follow-on error after a content-filter response, rather than the selected model being unavailable. The CLI should ideally surface the original cause or recover without reporting a model support error.
The error shown is: Execution failed: 400 The requested model is not supported. (Request ID: 6231:23F1C:1A9683F:1CCEB35:6AE5217E)
Request ID: 6231:23F1C:1A9683F:1CCEB35:6AE5217E
Environment: Copilot CLI 1.0.105, Windows; Auto, Claude Opus 5.5, and other models
Affected version
1.0.105
Steps to reproduce the behavior
Type a prompt that that gets filtered for cybersecurity risk
Expected behavior
Error message should clearly state that the request was filtered. This is what used to hapen.
Update: I think this probably happened in part because the agent was spawning subagents with different models that had different cybersecurity policies. Changing my subagent policy fixed it. But the misleading error message should still be fixed.
Activity
mlindgren commentedon Oct 31, 2026
Update: I think this probably happened in part because the agent was spawning subagents with different models that had different cybersecurity policies. Changing my subagent policy fixed it. But the misleading error message should still be fixed.