Skip to content

Changelog

New updates and improvements to OpenCode

  1. v1.18.11

    Stopped MCP SSE connections from getting stuck in reconnect loops after server error responses

    Fixes

    • CoreFixed provider model configs that use interleaved reasoning fields like reasoning_text or custom field names.
  2. v1.18.10
  3. v1.18.9
  4. v1.18.8

    Improved compatibility with newer MCP servers and OAuth flows

    Fixes

    • CoreReconnects MCP servers after expired SDK sessions, including concurrent requests.
    • CoreHonors configured MCP OAuth callback ports in mcp debug.
    • CoreStops sending deprecated sampling defaults to newer Gemini models.
  5. v1.18.6

    Fixes

    Fixes

    • CoreFixed branch-specific repository caches so refreshing one reference no longer moves another branch checkout.
  6. v1.18.5

    Improve Claude adaptive thinking handling across more response shapes

    Fixes

    • CoreAvoid OpenAI Responses phase handling that could break some conversations.
    • CorePreserve grep symlink paths in search results.@remixz
    • CorePreserve Mistral reasoning history across turns.
    • CoreStabilize Mistral prompt caching.
    • CoreUse the correct prompt cache keys for each SDK.
    • CoreFix MiniMax M3 thinking variant selection.
  7. v1.18.4

    Reduce OpenAI provider header timeouts during slow connection setup

    Improvements

    • CoreUse adaptive thinking controls for Kimi models on Anthropic-compatible providers, with summarized reasoning output by default.@chouqin

    Fixes

    • CoreRespect provider-defined reasoning options instead of falling back to the wrong reasoning controls.
    • CoreRestore Azure Cognitive Services endpoint support for Azure-hosted models.
  8. v1.18.3
  9. v1.18.2

    Improved default reasoning depth for Meta models

    Fixes

    • CoreStopped subagents from launching nested subagents by default, with a configurable subagent_depth limit when needed.
  10. v1.17.20

    Updated Azure AI support for GPT-5.6

    Fixes

    • CoreRemoved an obsolete Codex workaround that could interfere with OpenAI Luna Responses Lite requests.