Simple, Transparent Pricing

Token usage includes both input (prompt) tokens and output (completion) tokens. We use the same tokenization as the model you deploy, so counts match what you'd see locally. You're only charged for successful requests.

Shared GPU tiers run on multi-tenant infrastructure where resources are shared across users. This is cost-effective but may have variable latency. Dedicated GPU gives you reserved hardware with guaranteed performance and isolation - ideal for production workloads requiring consistent latency.

No, you cannot change your endpoint tier. However, you can deploy new endpoints with different tiers.

Requests exceeding your tier's rate limits will receive a 429 (Too Many Requests) response with a Retry-After header. We recommend implementing exponential backoff in your client. Consider upgrading to a higher tier if you consistently hit limits.

No minimum commitment required for any tier. Pay-as-you-go pricing means you only pay for what you use. You can stop or delete endpoints at any time with no penalty.