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
Opencode GO deepseek V4.1 Flash is very slow #57040
The model has been working for an hour now with only 600k tokens used. This is unacceptable slow performance. Is there a way to speed it up that I do not know about?
I am on windows 11 using the opencode v2 dekstop app
Thanks for the report, and sorry it has been so slow. I think this is a duplicate of #51010, which tracks slow output from deepseek-v4.1-flash on OpenCode Go. Measurements there show its speed on the Go route has dropped well below launch levels and below the same model served elsewhere. That points to the Go service rather than anything you can change in the app, so there's no client setting that fixes it. Choosing a lower reasoning effort for the model, if your setup offers one, can cut the number of tokens it generates per step.
Possible duplicates and related issues, ordered by likelihood:
The bot will close this issue as a duplicate in 1 day unless you reply explaining how your problem is different. For example, reply if every model is slow for you, not only DeepSeek V4.1 Flash, or if the session stops producing output entirely instead of just generating slowly. If so, please include your OpenCode version and a rough idea of how long a single reply takes.
Closing this as a duplicate of #51010, since we haven't heard that your case is different. That issue covers the same slowdown: deepseek-v4.1-flash on OpenCode Go generating much more slowly than at launch and than on other routes. Please follow #51010 for updates.
If other models are slow for you too, or a session stops producing output entirely, please open a new issue. Include your OpenCode version and roughly how long a single reply takes.
Activity
opencode-agent commentedon Oct 19, 2026
Thanks for the report, and sorry it has been so slow. I think this is a duplicate of #51010, which tracks slow output from
deepseek-v4.1-flashon OpenCode Go. Measurements there show its speed on the Go route has dropped well below launch levels and below the same model served elsewhere. That points to the Go service rather than anything you can change in the app, so there's no client setting that fixes it. Choosing a lower reasoning effort for the model, if your setup offers one, can cut the number of tokens it generates per step.Possible duplicates and related issues, ordered by likelihood:
deepseek-v4.1-flash) on the same Go route, much slower than on other routes and slower than at launch (open, with up-to-date measurements)The bot will close this issue as a duplicate in 1 day unless you reply explaining how your problem is different. For example, reply if every model is slow for you, not only DeepSeek V4.1 Flash, or if the session stops producing output entirely instead of just generating slowly. If so, please include your OpenCode version and a rough idea of how long a single reply takes.
opencode-agent commentedon Oct 20, 2026
Closing this as a duplicate of #51010, since we haven't heard that your case is different. That issue covers the same slowdown:
deepseek-v4.1-flashon OpenCode Go generating much more slowly than at launch and than on other routes. Please follow #51010 for updates.If other models are slow for you too, or a session stops producing output entirely, please open a new issue. Include your OpenCode version and roughly how long a single reply takes.