whichdevtool?
Menu
StacksDirectory Pricing changes Glossary
VPS and cloud servers App hosting and PaaS Managed databases Transactional email APIs Email marketing and newsletters Domain registrars CDN, DNS and edge security Monitoring, uptime and error tracking Auth and identity Frontend and full-stack frameworks Backend frameworks UI component libraries AI model APIs File and object storage Payments and billing Background jobs, queues and workflows ORMs and database toolkits

Inference provider (open-weight models) · by Cerebras

Cerebras Inference

Inference API on Cerebras wafer-scale chips for a short list of open-weight models, prepaid by credit with no permanent free tier.

Visit cerebras.ai ↗ Alternatives Trial
Cerebras Inference homepage
§1

Overview

Cerebras builds wafer-scale AI chips, and Cerebras Inference is its API for running open-weight models on them. The public catalogue is short: gpt-oss-120b, Z.ai GLM 4.7, Gemma 4 31B (with image input) and Qwen 3.8 27B, with more model families only on dedicated endpoints. Its docs list gpt-oss-120b at about 3,000 tokens per second.

There is no permanent free tier. New accounts receive $5 of credit after adding a verified payment method, valid for 30 days, with trial limits of 5 requests per minute and 1M tokens per day per model. The Developer tier is prepaid: you buy credits, which expire after a year, and when the balance hits zero your rate limits revert to zero until you add more or auto-recharge fires.

Developer limits are high (1M tokens per minute and 1,000 requests per minute on gpt-oss-120b, with no daily caps), and cached tokens only count toward a separate total-token bucket, so prompt caching lifts throughput.

§2

Pricing and plans

Plan Price What you get
Free Trial $5 credit After adding a verified payment method; expires after 30 days; 5 RPM, 30K TPM, 1M tokens/day per model
Developer (pay as you go) Prepaid credits gpt-oss-120b at 1M TPM and 1,000 RPM; no hourly or daily caps
gpt-oss-120b $0.35 in / $0.75 out per 1M About 3,000 tokens/sec; 131K context on paid
Enterprise Custom Provisioned limits, dedicated endpoints for additional model families

Read 2026-09-29. The cerebras.ai/pricing page renders its price tables with JavaScript and showed no figures to our fetch; the gpt-oss-120b price comes from Cerebras's own model documentation page. Per-token prices for GLM 4.7, Gemma 4 and Qwen 3.8 could not be confirmed. Purchased credits expire one year after purchase; auto-recharge can be set with a threshold and top-up amount. Rate limits use a dual bucket: an uncached tokens-per-minute limit and a total limit of 3x that, so cache hits raise effective throughput. Cerebras models are also available through AWS Marketplace, OpenRouter, Hugging Face and Vercel.

Source: inference-docs.cerebras.ai/models/openai-oss · prices as published, excluding tax.

§3

What it does

Wafer-scale hardware
Models run on Cerebras's own chips; gpt-oss-120b is listed at about 3,000 tokens/sec.
Prepaid credits
Credits last one year; auto-recharge triggers at a balance threshold you set.
Cache-aware limits
Separate uncached and total (3x) token-per-minute buckets reward cache hits.
Partner access
Also reachable through AWS Marketplace, OpenRouter, Hugging Face and Vercel.
Dedicated endpoints
Enterprise option for additional model families and custom limits.
§4

Strengths and limitations

Strengths

  • High published output speed on gpt-oss-120b.
  • No hourly or daily caps on the Developer tier.
  • Prepaid model means spend stops at your balance.

Limitations

  • No permanent free tier; the $5 trial requires a payment method and expires in 30 days.
  • Only four public models in the self-serve catalogue.
  • API access stops (limits drop to zero) when credits run out unless auto-recharge is set.
  • Free-trial context is capped at 65K tokens on gpt-oss-120b versus 131K on paid.
§5

Who it suits

Good for

  • Real-time agents and chat where output speed is the priority
  • Teams already using gpt-oss-120b who want faster responses

Look elsewhere if

  • Projects that need a broad model catalogue
  • Hobby use that needs an indefinite free allowance
§6

Alternatives to Cerebras Inference

Tool From Free option Stages
Together AI Inference provider (open-weight models)
Usage-based Paid
Claude API AI model API (model maker)
Usage-based Trial
OpenAI API AI model API (model maker)
Usage-based Paid
Mistral API (Mistral Studio) AI model API (model maker)
Usage-based Free tier
Fireworks AI Inference provider (open-weight models)
Usage-based Trial
OpenRouter Model router (unified API)
— Free tier

All AI model APIs →

§7

Questions

Does Cerebras have a free API tier?

No permanent one. New accounts get $5 of credit after adding a verified payment method, expiring 30 days after it is granted.

How much does gpt-oss-120b cost on Cerebras?

Cerebras's model docs list $0.35 per 1M input tokens and $0.75 per 1M output tokens on 2026-09-29.

What happens when my Cerebras credits run out?

Your rate limits revert to zero until you add credits; automatic recharge can prevent the interruption.

Sources read for this page