Skip to content

fix(opencode): restore reasoning variants for openai-compatible deepseek glm and minimax#23335

Open
elonazoulay wants to merge 1 commit intoanomalyco:devfrom
elonazoulay:fix/opencode-openai-compatible-reasoning-variants
Open

fix(opencode): restore reasoning variants for openai-compatible deepseek glm and minimax#23335
elonazoulay wants to merge 1 commit intoanomalyco:devfrom
elonazoulay:fix/opencode-openai-compatible-reasoning-variants

Conversation

@elonazoulay
Copy link
Copy Markdown

@elonazoulay elonazoulay commented Apr 18, 2026

Issue for this PR

Closes #23334

Type of change

  • Bug fix
  • New feature
  • Refactor / code improvement
  • Documentation

What does this PR do?

Restores low / medium / high reasoning variants for deepseek, glm, and minimax models routed through @ai-sdk/openai-compatible.

On current dev, those model families are caught by the same blanket exclusion guard and incorrectly return {} instead.

This change is intentionally narrow: it keeps the current exclusions for qwen, kimi, k2p5, mistral, and big-pickle in place. It is a follow-up to PR #21212, not a blanket revert.

How did you verify your code works?

From packages/opencode:

  • bun test test/provider/transform.test.ts
  • bun typecheck

Added focused coverage for:

  • deepseek on @ai-sdk/openai-compatible returning low / medium / high
  • glm on @ai-sdk/openai-compatible returning low / medium / high
  • minimax on @ai-sdk/openai-compatible returning low / medium / high
  • deepseek staying excluded on a non-compatible SDK

Screenshots / recordings

If this is a UI change, please include a screenshot or recording.

Checklist

  • I have tested my changes locally
  • I have not included unrelated changes in this PR

@elonazoulay elonazoulay force-pushed the fix/opencode-openai-compatible-reasoning-variants branch from 2c4b5ae to 3ef0d34 Compare April 18, 2026 21:03
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Bug: openai-compatible deepseek/glm/minimax models lost reasoning variants after qwen exclusion

1 participant