Use case
Picking a model per request
Send one request and let a router choose which generator should answer it.
What matters
You are buying a choice, not a label. Price is variable. You should still know which model ran if you need an audit trail.
latency · cost · accuracy
Top picks
- #1TypeSafe: Jev Router
Built to pick a model and a reasoning effort. It is not a classifier you threshold yourself.
Budget pick
No priced model in this shortlist.
Open-weight pick
None of these picks have a confirmed open-weight id.
Pitfalls
- Perplexity's Orchestrator API fails over between deployments of the model you named. It does not choose the model.
- Variable router pricing will not match a flat $/M on a single model.
Sample prompt
Send the user request to Jev Router when you want it to choose the generator. Log the model id it actually called.