[FEATURE]: Add disable_reasoning parameter for GLM/Cerebras thinking toggle #8441

Closed
opened 2026-02-16 18:09:58 -05:00 by yindo · 2 comments
Owner

Originally created by @srs-adamr on GitHub (Feb 3, 2026).

Originally assigned to: @thdxr on GitHub.

Problem

GLM-4.5/4.6/4.7 and Cerebras models support disable_reasoning boolean parameter to control thinking/reasoning output. Currently no way to configure this in OpenCode agent settings.

  • Issue #3259 asked "How to disable thinking in GLM-4.6?" but closed without implementation
  • Useful for reducing tokens, faster responses, avoiding parsing bugs (#6486, #7779, #6708)

Proposed Solution

Add disable_reasoning as optional boolean in agent configuration:

{
  "agent": {
    "fast-build": {
      "model": "cerebras/zai-glm-4.7",
      "disable_reasoning": true
    }
  }
}

Benefits

  • Reduce token usage/costs for simple tasks
  • Faster responses when deep reasoning not needed
  • Avoids reasoning tag parsing issues
  • User control over thinking verbosity

Implementation Notes

  • Apply to GLM and Cerebras providers
  • Default: false (thinking enabled)
  • Parameter name matches GLM/Cerebras API spec
Originally created by @srs-adamr on GitHub (Feb 3, 2026). Originally assigned to: @thdxr on GitHub. ### Problem GLM-4.5/4.6/4.7 and Cerebras models support `disable_reasoning` boolean parameter to control thinking/reasoning output. Currently no way to configure this in OpenCode agent settings. - Issue #3259 asked "How to disable thinking in GLM-4.6?" but closed without implementation - Useful for reducing tokens, faster responses, avoiding parsing bugs (#6486, #7779, #6708) ### Proposed Solution Add `disable_reasoning` as optional boolean in agent configuration: ```json { "agent": { "fast-build": { "model": "cerebras/zai-glm-4.7", "disable_reasoning": true } } } ``` ### Benefits - Reduce token usage/costs for simple tasks - Faster responses when deep reasoning not needed - Avoids reasoning tag parsing issues - User control over thinking verbosity ### Implementation Notes - Apply to GLM and Cerebras providers - Default: `false` (thinking enabled) - Parameter name matches GLM/Cerebras API spec
yindo closed this issue 2026-02-16 18:09:58 -05:00
Author
Owner

@iamadamreed commented on GitHub (Feb 3, 2026):

Closing - created under wrong account. Recreating as iamadamreed.

@iamadamreed commented on GitHub (Feb 3, 2026): Closing - created under wrong account. Recreating as iamadamreed.
Author
Owner

@github-actions[bot] commented on GitHub (Feb 3, 2026):

This issue appears to be a duplicate of #11987, which contains the same feature request for adding a disable_reasoning parameter for GLM/Cerebras models.

Both issues have identical titles and descriptions. Please consolidate discussion in #11987.

@github-actions[bot] commented on GitHub (Feb 3, 2026): This issue appears to be a duplicate of #11987, which contains the same feature request for adding a `disable_reasoning` parameter for GLM/Cerebras models. Both issues have identical titles and descriptions. Please consolidate discussion in #11987.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#8441