Feature request for product/service
AI Models
Describe the request
I recently ran some comparisons between Auto mode and manually selecting Sonnet 4.5 (full results here). Auto performed well, but the biggest gap in the test was that I couldn’t see which model it actually picked for each task.
Right now when you use Auto, there’s no indicator showing what model is running under the hood. You get the output but no visibility into the routing decision. I checked the GUI after a response and there’s nothing there either.
It would be useful to have a small label or indicator (even just in the response metadata or a tooltip) showing which model Auto selected. Something like “Auto → Sonnet 4.5” or “Auto → Opus” after the response comes back.
This would help with:
- Debugging: If Auto gives a bad response, knowing which model it picked helps you decide whether to retry with a different one or whether Auto just made a bad routing call
- Trust: A lot of the “Auto is cheating you” skepticism comes from the black box. Showing the routing would let people verify for themselves
- Learning: Over time you’d start to see patterns in what Auto picks for what, which helps you make better manual choices when you want to
The CLI has zero visibility into this. I haven’t found a way to see this in the GUI either, but I might be missing it.
Anyone else want this?