On-demand
Pay-as-you-go billing with no minimum commitment. GPU instances from USD 1.50 per GPU hour, models via Serverless Inference per million tokens.
No minimum commitment and no setup fees
Billed per hour or per second
No charge for network ingress and egress
GPU instances per GPU hour: NVIDIA L40S USD 1.50, A100 80GB PCIe USD 2.00, A100 80GB SXM USD 2.30, AMD MI300X USD 3.45, NVIDIA H100 USD 4.50, H200 USD 5.40, B200 USD 8.50, HGX B300 USD 9.90
NVIDIA GB200 NVL72 and AMD MI355X on request
CPU instances per vCPU hour: general-purpose USD 0.04, storage-optimized USD 0.09
Storage per GiB per month: object storage USD 0.06, shared disks USD 0.07, persistent disks USD 0.08, container registry USD 0.10
Managed Kubernetes USD 0.10 per cluster hour
Serverless Inference per 1M tokens from USD 0.05 input and from USD 0.20 output, depending on the model
Serverless Fine-Tuning per 1M training tokens from USD 0.40, depending on the model
Self-Serve Deployments per GPU hour: NVIDIA H100 USD 6.30, H200 USD 7.50, AMD MI355X USD 8.20, NVIDIA B200 USD 10.20, HGX B300 USD 13.80
