Claude Opus 4.6 returns 500 Internal Server Error in Copilot while GPT models work #293640
Description
Bug description
When using GitHub Copilot Chat with the Claude Opus 4.6 model, every request fails with a 500 Internal Server Error after ~5 seconds.
The exact same prompts work correctly and consistently when switching to any GPT model.
This makes Claude Opus 4.6 unusable in Copilot, while the rest of Copilot works normally.
Environment
IDE: VS Code
Extension: GitHub Copilot (latest version)
OS: Windows (10)
Copilot Chat enabled
Model affected: Claude Opus 4.6
Models not affected: all GPT models
Steps to reproduce
Open VS Code
Open Copilot Chat
Select Claude Opus 4.6 as the model
Send any medium or long technical prompt (for example a code analysis or architecture description)
After ~5 seconds, Copilot returns a 500 Internal Server Error
Expected behavior
Claude Opus 4.6 should respond normally, similarly to GPT models, without server errors.
Actual behavior
Copilot returns a 500 Internal Server Error for every request using Claude Opus 4.6.
Once this happens, the chat session often becomes unusable until a new conversation is started.
Additional notes
This issue started immediately after switching to Claude Opus 4.6
No changes were made to the local environment
The problem is fully reproducible
Switching back to a GPT model resolves the issue instantly