Repository navigation
[FEATURE]: Universal intelligent model & compute router (provider, quota, cost, quality, reasoning aware) #52743
Description
- I have searched the existing issues and confirmed this feature has not already been requested in the same form.
Description
OpenCode should offer a provider-agnostic Auto model mode that automatically selects the appropriate provider, model, and reasoning level for each task.
The main user problem is that users with multiple providers, subscriptions, APIs, and models currently need to manually decide which model to use for every task.
The router should consider:
- task complexity and required capabilities
- available providers/models
- reasoning level
- model/API cost
- subscription or quota state when this information is available
- user preferences such as quality vs. cost
Example:
- small edit → cheaper/faster model
- normal coding task → balanced model
- complex debugging → stronger model with higher reasoning
- preferred model unavailable or rate-limited → fallback to another allowed model
The user should be able to select something like:
/models → Auto
and still manually override the selected model at any time.
This overlaps with #51972, #51362, #34370 and #8456, but the additional request here is specifically for provider-agnostic routing that also considers cost, subscriptions/quota state, and user preferences rather than only task type.
If this fits better as an extension of #51972 rather than a separate feature, I am happy for this issue to be treated as additional requirements for that work.
Activity
github-actions commentedon Oct 2, 2026
This issue doesn't fully meet our contributing guidelines.
What needs to be fixed:
Please edit this issue to address the above within 72 hours, or it will be automatically closed.
This issue might overlap with existing issues. Please check:
If you believe this was flagged incorrectly, please let a maintainer know.
github-actions commentedon Oct 2, 2026
The issue still needs two updates to meet the feature template: add its required verification checkbox, and substantially condense the description to the core user need and requested behavior. The current detailed proposal remains too lengthy. Please keep the useful context while removing the extensive examples, configuration sketch, and repeated requirements.
mixxermannipro commentedon Oct 2, 2026
Updated. I added the required verification checkbox and condensed the issue to the core user need and requested behavior.
github-actions commentedon Oct 2, 2026
Thanks for updating the issue. The revised description and feature-template details are clear.
opencodejr commentedon Oct 28, 2026
Thanks for the detailed proposal! I think this is a duplicate of #34370, the open issue where maintainers have pointed the broader discussion of automatic, task-based model routing. Your cost, quota and preference criteria fit well as extra requirements there, and you've already linked this issue from it.
Possible duplicates and related issues, ordered from most to least likely:
automodel mode with a classifier; closed as a duplicate of [FEATURE]: opencode could automatically use different models based on task type #8456For now, you can get part of the way by giving each agent its own model, for example a cheaper model for the
exploresubagent and a stronger one for your main agent: https://opencode.ai/v2/docs/agents#modelThe bot will close this issue as a duplicate in 1 day unless you reply explaining how your request is different from #34370.
opencodejr commentedon Oct 29, 2026
Closing this as a duplicate of #34370, since we haven't heard back in the past day. That issue is still open and is where maintainers have pointed the discussion of automatic, task-based model routing. Your additional criteria (provider-agnostic selection, cost, quota/subscription state, user preferences and fallback on rate limits) would be a useful addition there.
For now, you can give each agent its own model, for example a cheaper one for the
exploresubagent: https://opencode.ai/v2/docs/agents#modelIf you think this request is different from #34370, reply here and we can reopen it. Thanks again for the detailed proposal!