We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
Pricing is quote-based.
Cerebras Inference has no free plan, but the Developer plan includes a one-time $5 promotional credit with a valid payment method. It expires after 30 days, and Developer is for development, evaluation and experimentation—not production. After the credit, you pay per token: rates differ by model and between input and output tokens. Enterprise capacity is custom-priced. PayGo credit prices and training add-on prices are not published, so production buyers should request a quote before estimating total cost.
01Every plan, line by line
| Plan | Billed annually | Billed monthly | Built for |
|---|---|---|---|
| Developer | — | Quote | Development, evaluation, and experimentation |
| Enterprise | — | Quote | Production workloads needing tailored capacity |
Add-ons and overages
- PayGo credits; prices not published
- Customized model weights, fine-tuning, and training by agreement; prices not published
02What the plan grid doesn't tell you
Production capacity
Enterprise pricing is custom. Ask for a quote that states capacity, priority, rate limits, latency options and supported models.
PayGo credits
PayGo credits are listed as an option, but their prices are not published. Confirm credit cost, validity and how usage is deducted.
Training and weights
Customized model weights, fine-tuning and training are available by agreement; the vendor does not publish their prices.
Token mix
Input and output rates differ by model. Estimate both token types against expected traffic; the listed rates do not give a fixed per-request cost.
03Which plan should you buy?
| If you are… | Get | Why |
|---|---|---|
| Side project or evaluation | Developer | Use pay-as-you-go rates for development and evaluation; the $5 credit expires after 30 days. |
| Startup in production | Enterprise | Developer is not intended for production; Enterprise offers production-ready capacity at custom pricing. |
| Team needing tailored capacity | Enterprise | The plan lists best priority, higher rate limits and enhanced latency options; ask for a workload-specific quote. |
04Is Cerebras Inference worth the money?
Developer can be a practical way to price an evaluation: the listed token rates are clear by model and input or output, and the one-time $5 credit lowers the initial bill. It is not a free tier and is explicitly not intended for production. Enterprise is the production option, but its custom price makes a quote necessary before comparing total cost. Budget separately for any agreed training or model-weight work, and ask about rate limits and capacity in the quote.
Cerebras Inference pricing FAQ
How much does Cerebras Inference cost?
Developer usage is billed per token. GPT OSS 120B is listed at $0.35 per 1M input tokens and $0.75 per 1M output tokens. Qwen 3.8 27B is $0.99 per 1M input tokens and $1.49 per 1M output tokens. Enterprise pricing is custom.
Does Cerebras Inference have a free tier?
No. The pricing lists a one-time $5 promotional credit with a valid payment method, expiring after 30 days. It is not a recurring free tier. Developer usage is pay-as-you-go, and the plan is intended for development, evaluation and experimentation.
Is Cerebras Inference Enterprise pricing public?
No. Enterprise is custom-priced. The plan lists production-ready capacity, best priority, higher rate limits and enhanced latency options. Ask Cerebras for a quote that specifies capacity, model access, limits and any deployment requirements.
Are fine-tuning and training included in the token rates?
The pricing lists customized model weights, fine-tuning and training as services by agreement, with prices not published. Do not assume these services are covered by inference token rates; request a separate quote and confirm the scope.