Published Inference Rates, by Token
and by GPU Hour
From flexible serverless inference to reserved private capacity, choose the operating model that fits your performance, control, and cost requirements.
Serverless Inference
Per-token rates on shared capacity. Pay for what you send and receive.
See PricingData Agent Factory
Scoped per dataset. Cost follows data volume and the quality gate you set.
HOW IT"S SCOPEDDedicated Endpoints
Hourly rates on GPUs reserved for you. Isolated capacity, steady throughput.
TALK TO SALESModelInput / outputGPU / hour