Skip to content

Pricing

Published rates. No sales call required.

Start free and grow at your own pace, paying only for what you use. Dedicated GPU capacity is quoted per configuration. No fees to switch or leave.

Inference

Rates per model.

USD per 1M tokens · As of September 2026

Inference pricing per million tokens
AttributeContextInputCached inputOutput
DeepSeek V4.1 FlashVISIONdeepseek/deepseek-v4.1-flash1M$0.20$0.006$0.60
DeepSeek V4 Prodeepseek/deepseek-v4-pro1M$1.00$0.10$2.00
GLM 5.2z-ai/glm-5.2256K$1.00$0.15$3.00
Kimi K3VISIONmoonshotai/kimi-k31M$2.85$0.29$14.25

Moving a call to a cheaper model is a string change, not an integration. Live rates are also published at the models endpoint.

Public cloud

Self-serve or dedicated.

Aurora Portal

Storage & compute

$5.99

per TiB / month stored

Compute in Early Access

  • S3-compatible API, no request fees
  • Versioning and Object Lock
  • VMs, volumes, and snapshots
  • Private networks and public IPs
Tailored

Dedicated capacity

Quote

per configuration

OPEX, no hardware to buy

  • NVIDIA B200, B300, GB300 NVL72
  • Single-tenant clusters
  • SLA and support terms per contract
  • Live in weeks

Billing FAQ

The fine print, in large type.

Yes. For 30 days from sign-up you get 5 GiB of object storage, one bucket, five access keys, and access to the Inference API, with no payment method required. Compute isn’t part of the trial. One trial per customer.

Open the portal. See it for yourself.

Create an account, make a bucket, and send your first inference call in minutes. No sales call, no commitment.

30-day free trial · no payment method required