Explore inference routes by workload, provider, and model.
Browse workload, provider, and model combinations. Check current endpoint availability and evaluate the candidate on your own tasks before using a route.
495 pages covering use case × provider × model combinations from the reference catalog. Verify current availability with the provider.
Reference combinations are a starting point for evaluation, not a guarantee of availability or savings.
495 reference combinations. Availability can change; confirm the endpoint before deployment.
Paste a week of traffic. Get the number that books the audit.
See what you're overpaying →