Large orchestration prompts, small/simple prompts, planning, no matter it always selects Grok 4.5. Tried changing to optimize for intelligence, balance, cost… all Grok. Seems interesting to call this an auto router when it only selects the model of the company that is aquiring you.
Hey, thanks for the report. This isn’t a router bug. I’ll break it down.
Auto also called Cursor Router or Premium routing is a dynamic model choice per request from the pool available on your plan. It’s based on internal benchmarks, quality scores, and current load capacity, so the pool isn’t fixed and changes over time. Grok 4.5 often gets picked right now. That’s part of dynamic routing based on task fit and capacity, not bias toward any model. Model choice is tied to performance and how well it matches the task.
About the optimize for Cost, Balance, Intelligence modes. That’s part of Cursor Router, and right now it’s only available on Teams and Enterprise. Individual and free plans don’t have it yet. So switching those options on your plan doesn’t actually affect routing today.
If you want predictability and a specific model, don’t use Auto. Pick a model manually in the model picker. Then you’ll always know what’s running under the hood.
If you want us to check your specific requests, send the Request ID Chat menu top right > Copy Request ID. If Privacy Mode is on, it might be unavailable. Also tell us which email you’re logged in with. The Request ID shows which model was actually selected.
I’m using an enterprise account and no matter what I choose Cost, Balance, or Intelligence I always get Grok 4.5 High. Everyone in the company is hitting this. And it’s destroying our limited budget, going through it much faster than the old auto mode.