CoPilot Error, Help anyone?

[booxpro012]

Hey all, since the latest VS code update, I've been getting having error messages sent back from my model requests.

Reason: Request Failed: 400 {"message":"You invoked an unsupported model or your request did not allow prompt caching. See the documentation for more information."}

Anyone else have this issue? Strange, nothing has changed. Checked my settings, everything appears in order.

Thanks to all who read.

[ganapathijahnavi]

Yes, a few people have run into this after recent updates. That error usually means either:

1. The model you’re using was deprecated/renamed, or
2. The request is sending prompt caching, but the model doesn’t support it.
   Even if you didn’t change settings, the recent update may have changed defaults. Try switching to a different model or disabling prompt caching (if there’s an option), then test again.

[booxpro012]

The models im using are Claude opus 4.6, and Claude sonnet 4.6, these would not be depreciated models, unless because they are newly released they are having issues? I’ve tried using various Claude models that are still in use but no go.  There is no cache disabling option I could locate. All my research points to a backend connectivity issue on GitHub / anthropic endpoints. Only Claude models are having this issue.  It's strange, though, my GPT 5.3 codex models are disconnecting in mid call also. In any case, thank you for your reply. Hopefully I can narrow it down.

[midiakiasat]

400 with:

"You invoked an unsupported model or your request did not allow prompt caching"

means the model ID being sent by the extension is not accepted by the provider endpoint or the extension is sending prompt-caching flags to a model that does not support them.

Key points:

1. Model name mismatch
   VS Code updates sometimes change internal model IDs.
   “Claude opus 4.6” / “Claude sonnet 4.6” in UI does not guarantee the backend ID is valid.
   This is usually a provider-side alias change.

2. Prompt caching incompatibility
   Some Claude endpoints do not support prompt caching.
   If the extension now enables caching by default, the request fails with 400.

3. Mixed failure (Claude + GPT mid-call)
   If GPT calls are disconnecting mid-stream, that suggests:

   • Extension version regression
   • Corrupted auth/session token
   • Provider routing instability
     Not a local config error.

Immediate actions:

• Sign out of GitHub Copilot in VS Code.
• Remove ~/.config/github-copilot (or %APPDATA%\GitHub Copilot on Windows).
• Restart VS Code.
• Re-authenticate.
• Explicitly re-select a stable model (avoid newest aliases).

If still failing:

• Downgrade VS Code to previous working version.
• Downgrade Copilot extension to previous version.
• Check DevTools → Network to confirm actual model ID being sent.

If only Claude models fail and GPT works intermittently, this is almost certainly a backend model routing or capability flag mismatch introduced in the latest extension release, not your configuration.

[booxpro012]

Thank you, that was extremely helpful and fixed my issue.
Very much appreciate the help.

[shinybrightstar]

Welcome to the GitHub Community, @booxpro012, we're happy you're here! You are more likely to get a useful response if you are posting your question(s) in the applicable category. Check out this Guide on how to post in the right category. It'll ensure your questions and discussions reach the right people faster!

This particular discussion belongs in the Copilot Conversations category, I've gone ahead and moved it for you. Good luck!

[edd426]

The fix for me was setting github.copilot.chat.anthropic.useMessagesApi to false in user settings. The Messages API path sends prompt-caching headers that fail behind corporate proxies.