Home/Large language models/Cerebras Inference

Cerebras Inference pricing

Cerebras sells fast AI inference on its own wafer-scale chips through a pay-as-you-go API and enterprise contracts.

Large language modelsPriced per M input tokensSelf-serve + sales
Start 30-day trialPrices checked Nov 8, 2026

How much does Cerebras Inference cost?

Cerebras Inference charges by usage, from $0.35 per M input tokens. 5 plans listed. No price change since Q3 2025, across 4 archived reads.

Usage price from $0.35 per M input tokens
Free plan
No
Free trial
One-time $5 promotional credit, expires 30 days after activation
Usage price
$0.35 per M input tokens
Enterprise
Production workloads, all supported models, dedicated account team, priority and higher rate limits, quote-based pricing"Contact us for a quote"
Read from www.cerebras.ai/pricing on Nov 8, 2026. Price changed? Tell usSee the line each price came from

6 other tools in this category publish a price, from $8.

Compare all 32 large language models

In your favor

  • No overage charges. Not automatically charged; access pauses when credit is exhausted unless PayGo credits are purchased
  • Free trial. One-time $5 promotional credit, expires 30 days after activation

Watch out for

  • Service stops at the limit. Not automatically charged; access pauses when credit is exhausted unless PayGo credits are purchased

How you buy and get started

Self-serve + salesButtons on their site: GET API KEY, CONTACT SALES, GET STARTED
Sign up
Self-serve API key signup for the Developer tier
"Self-serve pay-as-you-go with free $5 credit to start"
Setup
Get started in under 30 seconds
"Get started in <30 seconds with drop-in OpenAI API compatibility."
When sales steps in
Enterprise production workloads are sold by quote
"Contact us for a quote"
Support
Community Discord (Developer)
"Community Discord"
Support
Dedicated account team (Enterprise)
"Dedicated account team"

What Cerebras Inference does

Cerebras runs open-weight and partner models on its own hardware, offered through a self-serve Developer tier billed per token and an Enterprise tier sold by quote. It also offers fine-tuning and training services by agreement, and on-premise systems.

Good for
Developers who want low-latency inference on open models via an OpenAI-compatible API, and enterprises needing production capacity, dedicated support, or custom model weights.
Channels
API, Playground, AWS Marketplace, OpenRouter, Hugging Face, Vercel, On-premise
Main product
Cerebras Inference Cloud
Also includes
Developer tier: GPT OSS 120B and Qwen 3.8 27B, Enterprise tier: all supported models, Community Discord support (Developer), Dedicated account team (Enterprise), Fine-tuning and training services by agreement (Enterprise), One-time $5 promotional credit
Connects to
AWS Marketplace, OpenRouter, Hugging Face, Vercel, LiveKit, Oracle Cloud Infrastructure
Public API
Yes
  • Fine-tuning supportFine-tuning and training services
  • Tool and function callingNot stated
  • SOC 2 complianceSOC 2

Security and your data

QuestionCerebras Inference's answerWhere it says so
Trains AI on your data?no"the foregoing does not grant Cerebras the right to use Service Content for the purpose of training or fine-tuning models."terms
SOC 2listed on Trust Center"SOC 2"security
GDPRlisted on Trust Center"GDPR"security
CCPAlisted on Trust Center"CCPA"security
Trust centerhttps://trust.cerebras.ai/"Welcome to Cerebras Systems' Trust Center"security

Not stated on the 6 pages we read (homepage, about, terms, privacy, security, pricing): HIPAA (signs a BAA), PCI, ISO 27001, Single sign-on.

Every plan

PlanPriceBilled yearlyIncludedOver the allowanceWhere it says so
Developer: GPT OSS 120Bper M input tokens $0.35per M input tokens – – – "$0.35/M tokens"
Developer: GPT OSS 120B outputper M output tokens $0.75per M output tokens – – – "$0.75/M tokens"
Developer: Qwen 3.8 27Bper M input tokens $0.99per M input tokens – – – "$0.99/M tokens"
Developer: Qwen 3.8 27B outputper M output tokens $1.49per M output tokens – – – "$1.49/M tokens"
Enterprise Quote – – – "Contact us for a quote"

Cerebras Inference price history

No price changes on Cerebras Inference's pricing page since 2025Q3, across 4 archived reads.

From archived copies of the pricing page in the Internet Archive, one per quarter, in USD.

Fine print

Billed beyond usage
Not automatically charged; access pauses when credit is exhausted unless PayGo credits are purchased
"You will not be charged or automatically enrolled in paid usage."

Not stated on the 6 pages we read (homepage, about, terms, privacy, security, pricing): Billing increment, Contract, Refunds, Auto renew, Price change notice.

About the company

Legal name
Cerebras Systems Inc.
Address
1237 E. Arques Ave, Sunnyvale, CA 94085
Domain registered
Dec 15, 2017
Founder
Andrew Feldman
Funding
Began trading on Nasdaq under ticker CBRS in May 2026
"In May, Cerebras begins trading on Nasdaq under the ticker CBRS."
Customers
"Battle-tested at scale by leading cloud service providers and enterprises."
their claim
Backing
"Investors include Sam Altman, Andy Bechtolsheim, Greg Brockman and others"
their words

Cerebras Inference alternatives

Tools that do the same job (LLM API), wherever they're listed. All 29 Cerebras Inference alternatives.

Reviews elsewhere

Questions

Is Cerebras Inference free?

No ongoing free plan. Developer tier is pay-as-you-go with a one-time $5 credit that expires after 30 days; access pauses when the credit runs out unless PayGo credits are purchased.

Does Cerebras Inference train AI on your data?

No, according to Cerebras Inference: "the foregoing does not grant Cerebras the right to use Service Content for the purpose of training or fine-tuning models."