impulse

Hosts

140 places to call a model. The same model can cost very different amounts depending on where you call it; each host page shows its price against the lab’s own.

Lab APIs

Each lab's own API. New models usually land here first, and its price is the headline others are compared with.

Clouds

AWS, Azure, Google Cloud and others resell lab models under their own contracts, regions and quotas.

Inference providers

GPU clouds that serve open-weight models, often cheaper or faster than the lab.

Gateways

One key for many providers. They route each request and may add a margin.

HostFrontier modelsAll models
5497
3998
2571
213
-22
-16

Subscription plans

Flat monthly plans that include model use, so they have no per-token price.

HostFrontier modelsAll models
4650
2525
11

Local runtimes

Run open weights on your own hardware.

HostFrontier modelsAll models
-29
-5