Start free, scale on usage-based pricing, and reserve dedicated capacity only when you need it. No minimums, no seat fees, no lock-in.
For prototypes and side projects.
For production apps and teams.
billed annually + usage
Start buildingFor scale, security, and compliance.
A sample of compute rates, priced per 1M tokens unless noted. The full price list lives in the model library.
You are billed only for what you use, metered per input and output token (or per image, per second of audio, or per embedding). There are no minimums, no seat fees, and no idle charges on on-demand compute.
On-demand bills per token with autoscaling and scale-to-zero — ideal for spiky or early-stage traffic. Dedicated reserves GPU capacity at a flat hourly rate for predictable, high-volume workloads with no per-token cost.
Yes. Committed-use discounts kick in automatically as monthly spend grows, and enterprise agreements can include custom rates, private models, and dedicated regions.
Every new account starts with free credits so you can build and test before you pay. No credit card required to start.
Prompts and completions are never used to train models. Enterprise plans add SOC 2 Type II, zero data retention, SSO/SAML, and single-tenant deployment options.
Create an account, grab free credits, and run your first model in under a minute.