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
[FEATURE]: Ability to specify different model to use for compaction #6976
I have verified this feature I'm about to request hasn't been suggested before.
Describe the enhancement you want to request
It would be nice to be able to use a lightweight model (eg flash / haiku) to handle context compression rather than the slower and more capable model that is currently running. Perhaps either as a config option or as a different agent like 'Compactor' which could then have a default model associated with it?
These issues address similar use cases around flexible model selection for different agent tasks. You may want to reference them or consolidate the discussion.
Per mzealey's note, is there any actual documentation on how to manage this per model? Some models have become unusuable for me due to aggressive compaction that doesn't appear to respond to the config.json options I've been able to find through various issue reports.
Activity
github-actions commentedon Jan 5, 2026
This issue might be a duplicate of existing issues. Please check:
These issues address similar use cases around flexible model selection for different agent tasks. You may want to reference them or consolidate the discussion.
rekram1-node commentedon Jan 5, 2026
@mzealey you can do this in ur opencode config:
mzealey commentedon Jan 5, 2026
Ah I see - I looked in the docs which did not have this - i see it's in the config.json though.
98h398hrpohpoai commentedon May 4, 2026
Per mzealey's note, is there any actual documentation on how to manage this per model? Some models have become unusuable for me due to aggressive compaction that doesn't appear to respond to the config.json options I've been able to find through various issue reports.