You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
desktop: model tooltip shows "No reasoning" for openai-compatible models configured with reasoning: true #51846
In the desktop app, the model picker tooltip reports "No reasoning" for custom openai-compatible models that are explicitly configured with reasoning: true and have working reasoning variants. Reasoning itself works fine at runtime (variants apply and thinking output is returned) — only the tooltip's reasoning check is wrong.
Root cause (from the bundled v2.0.18 renderer): the tooltip reads capabilities.reasoning, but the backend never populates a reasoning key inside capabilities — config merge only copies tools/input/output, so capabilities.reasoning is always falsy and the legacy model.reasoning fallback is never reached.
Possible fixes: populate capabilities.reasoning from the model's reasoning config field when building/merging models, and/or have the tooltip fall back to model.reasoning || model.variants?.length > 0.
Plugins
None
OpenCode version
v2.0.18 (desktop app + CLI, channel: latest)
Steps to reproduce
Define a custom openai-compatible provider in ~/.config/opencode/opencode.jsonc with a model configured with "reasoning": true and reasoning variants:
Thanks for the detailed report! The incorrect “No reasoning” tooltip despite working reasoning variants is the same problem as #50257, where we're tracking both built-in and custom OpenAI-compatible models. Please subscribe or add a 👍 there, and feel free to share any additional configuration details.
Activity
github-actions commentedon Sep 28, 2026
Thanks for updating the issue with the required description, reproduction steps, and system details!
neriousy commentedon Oct 20, 2026
Thanks for the detailed report! The incorrect “No reasoning” tooltip despite working reasoning variants is the same problem as #50257, where we're tracking both built-in and custom OpenAI-compatible models. Please subscribe or add a 👍 there, and feel free to share any additional configuration details.