The CentML Platform operates on a credit-based billing system, where 1 CentML credit equals 1 USD. You can buy credits through the Platform by going to Account -> Wallet.

​
Serverless endpoint

Serverless endpoint usage is billed according to the total number of tokens generated and processed.

Model SizeCredits per 1M Tokens
DeepSeek-R1 MoE (671B)$3.99
Llama-3.3 (70B)$0.50
Qwen2.5 (32B)*$0.80
Qwen2.5-VL (7B)*$0.15
Llama-3.1 (8B)*$0.10
Llama-3.2 (3B)*$0.06
Phi-3.5*$0.12

*Coming Soon…

​
Dedicated deployments

Dedicated deployments are charged based on the type and duration of hardware used, following a per-minute billing system.

GPUCredits per GPU per hour
Nvidia L4 24GB0.30
Nvidia A10G 24GB0.30
Nvidia A100 40GB1.10
Nvidia H100 80GB2.50
Private Inference EndpointsCodex Example Repo