opencode 1.18.4: adaptive thinking for Kimi models, Azure endpoint fix, desktop prompt input rewrite
Kimi models on Anthropic-compatible providers now use adaptive thinking controls with summarized reasoning output by default, and provider-defined reasoning options are honoured instead of falling back to the wrong controls. Azure Cognitive Services endpoint support is restored for Azure-hosted models; the desktop app gets a rewritten v2 prompt input and review-panel work.
PUBLISHED2026-07-20
OBSERVED2026-08-11
AGE22d
SOURCES1
Core
- Adaptive thinking controls for Kimi models on Anthropic-compatible providers, summarized reasoning output by default
- Provider-defined reasoning options now respected instead of falling back to the wrong reasoning controls
- Azure Cognitive Services endpoint support restored for Azure-hosted models
- Reduced OpenAI provider header timeouts during slow connection setup
Desktop
- v2 prompt input rewritten for more reliable command, context, shell, attachment and history interactions
- Embedded terminal theme synced with the app theme
- Review panel: open file tabs stay aligned with the current diff view; improved resizing and sticky controls
- Fixes: command drafts preserved in the command menu; no session-view failure on missing message parent links; empty prompt text parts blocked; model variant selection accessibility restored; keybind tooltips on prompt input controls; deduplicated file entries in diff summaries; recovery window no longer touches destroyed windows
COMMUNITY
No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.