GitHub Copilot: `gpt-5.4` appears to be available but fails at runtime.

[chrisarcand]

Describe the bug

With the GitHub Copilot provider, gpt-5.4 appears to be available but fails at runtime.

goose discovers the model successfully, but when it is used, the request fails with an error saying the model is not accessible via /chat/completions. This makes the model appear supported during selection/configuration even though it is not usable through the current request path.

To Reproduce
Steps to reproduce the behavior:

1. Configure goose to use the GitHub Copilot provider
2. Select or set the model to gpt-5.4
3. Run:

   GOOSE_PROVIDER=github_copilot GOOSE_MODEL=gpt-5.4 GOOSE_DISABLE_KEYRING=1 goose run --text "Reply with exactly: pong" --no-session --quiet

4. See error:
   Request failed: Bad request (400): model "gpt-5.4" is not accessible via the /chat/completions endpoint.

Expected behavior

If gpt-5.4 is surfaced as an available GitHub Copilot model, it should work when selected. Otherwise, it should not be presented as available/selectable through the current integration path.

---

• OS & Arch: macOS arm64
• Interface: CLI
• Version: v1.28.0
• Extensions enabled: Default built-in extensions
• Provider & Model: GitHub Copilot - gpt-5.4

---

Additional context

I verified this is not just a bad model name:

• gpt-5.4 is returned in the GitHub Copilot model list
• gpt-5.4 works when sent to GitHub Copilot's /responses API
• gpt-5.4 fails when sent to GitHub Copilot's /chat/completions API
  So the issue appears to be that the model is discoverable, but not usable through the endpoint goose currently uses for the GitHub Copilot provider.