Repository navigation
GPT-5.5 fails because reasoningEffort is added automatically #26219
Description
Description
This causes requests to fail for setups where gpt-5.5 is routed through /v1/chat/completions, especially with Azure/OpenAI-compatible gateways.
Error:
Bad Request: Error code: 400 - Function tools with reasoning_effort are not supported for gpt-5.5 in /v1/chat/completions. Please use /v1/responses instead.
The issue is caused by this condition also matching gpt-5.5. Only gpt-5-chat is currently excluded. gpt-5.5 should also be excluded here.
input.model.api.id.includes("gpt-5") && !input.model.api.id.includes("gpt-5-chat")
Plugins
None
OpenCode version
v1.14.40
Steps to reproduce
- Configure OpenCode with an Azure/OpenAI-compatible provider using
gpt-5.5. - Route requests through
/v1/chat/completions. - Start a new OpenCode session with
gpt-5.5. - Send any prompt.
- The request fails because OpenCode automatically includes
reasoningEffort: "medium".
Expected: OpenCode should not send reasoningEffort for gpt-5.5 when using chat completions.
Actual: The request fails with a 400 error.
Screenshot and/or share link
Operating System
macOS 26.4.1
Terminal
Mac Terminal
Metadata
Metadata
Assignees
Labels
No labels
Activity
frederiknsgo commentedon May 7, 2026
I tested this locally by running a patched OpenCode build against an OpenAI-compatible provider using
gpt-5.5over/v1/chat/completions.Without the workaround, requests consistently failed with:
I then added this exclusion:
to the
gpt-5reasoning block inProviderTransform.options().After rebuilding and rerunning OpenCode, requests succeeded normally and
gpt-5.5worked correctly through the chat completions endpoint.frederiknsgo commentedon May 7, 2026
Another option would be to make reasoningEffort configurable. Currently it passes reasoningEffort under water even if not specified in the config