What is Cerebras Inference
Cerebras Inference serves open models on Cerebras systems through compatible APIs, offering free-trial, pay-as-you-go and dedicated-endpoint access.
Public endpoint pricing
As of September 17, 2026, the live public catalogue contains two non-retired models. OpenAI GPT OSS 120B costs US$0.35 per million input tokens and US$0.75 per million output tokens, with a 131,072-token context limit. Qwen 3.8 27B costs US$0.99 input and US$1.49 output, supports vision and has a 65,536-token context limit.
Both public models are available on free-trial and pay-as-you-go tiers subject to rate limits. Supported API features include streaming, tools, structured outputs and reasoning controls; Qwen also supports image input.
Dedicated endpoints
Dedicated access uses custom flat monthly pricing based on required input and output token throughput, with reserved capacity, additional model families, higher throughput and production SLAs. Cerebras documents flexible contract terms of three, six or twelve months.
Official sources: Cerebras live public model catalogue and prices, Cerebras model catalogue, Cerebras pricing tiers.