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
[BUG] OpenCode Go DeepSeek V4.1 Flash intermittently returns reasoning-only responses after model ID migratio #48419
After OpenCode Go changed the DeepSeek model ID from opencode-go/deepseek-flash to opencode-go/deepseek-v4.1-flash, I started seeing intermittent empty responses in existing sessions.
I reproduced this independently on two Linux servers.
When the problem occurs, OpenCode persists a reasoning part but no text part. The reasoning sometimes contains both the internal planning and what is clearly the complete final user-facing answer, concatenated together.
The turn then ends normally with finish: stop.
For example, the persisted structure is:
step-start
reasoning
step-finish
instead of the normal:
step-start
reasoning
text
step-finish
Restarting OpenCode or the server does not reliably fix it. However, after several more messages in the same session using deepseek-v4.1-flash, responses eventually return to normal without any configuration change.
This happened on both servers specifically after switching existing sessions from deepseek-flash to deepseek-v4.1-flash.
I have not captured the raw OpenCode Go SSE response, so I cannot tell whether the final content is already missing upstream or whether it is being classified as reasoning inside the OpenCode/OpenCode Go pipeline.
Plugins
None relevant to the issue.
OpenCode version
1.18.30
Steps to reproduce
Use an existing OpenCode session with opencode-go/deepseek-flash.
Switch the session/agent model to opencode-go/deepseek-v4.1-flash.
Continue chatting in the same session.
In some of the first several turns, the response is empty.
Inspect the persisted session data in ~/.local/share/opencode/opencode.db.
The affected assistant message contains a reasoning part but no text part and finishes with stop.
Continue sending messages without restarting or changing configuration.
After several turns, normal reasoning -> text responses eventually resume.
Observed independently on two Linux servers(Ubuntu 24.04).
Screenshot and/or share link
No public share link because the affected sessions contain private conversation history.
I can provide sanitized SQLite session records if needed.
If your issue is specifically about the model ID migration from deepseek-flash to deepseek-v4.1-flash triggering this behavior in existing sessions, that aspect may be distinct. Please check the above issues and link back if yours is a different root cause.
Activity
github-actions commentedon Sep 11, 2026
This issue might be a duplicate of existing issues. Please check:
reasoning_contentinstead ofcontentin streaming responses, causing the pipeline to classify text as reasoningreasoning_contentis dropped in tool call messagesIf your issue is specifically about the model ID migration from
deepseek-flashtodeepseek-v4.1-flashtriggering this behavior in existing sessions, that aspect may be distinct. Please check the above issues and link back if yours is a different root cause.thdxr commentedon Oct 11, 2026
Thanks for the report! OpenCode 2.0 is now out and development is focused on V2, so we're closing issues filed against OpenCode 1. Please upgrade (
opencode upgrade --major, details: https://anoma.ly/notes/opencode-2-0), and if this is still relevant on 2.0, reopen this issue or open a new one with your 2.0 version.— from 𝕺𝖕𝖊𝖓𝕮𝖔𝖉𝖊