Cerebras Pricing Calculator

The public Cerebras rate card lists two models as of August 2026: GPT-OSS-120B at $0.35 input/$0.75 output per million tokens (production) and Google Deepmind Gemma 4 31B at $0.99 input/$1.49 output (Preview). ZAI-GLM-4.7, previously listed at $2.25 input/$2.75 output, was removed from the public card on its disclosed deprecation date of August 17, 2026. Preview models are intended for evaluation only, so GPT-OSS-120B is the only public model sanctioned for production. Other models such as Llama 3.3 70B, Qwen3-32B, and now the Z.AI GLM family are available via Dedicated Endpoints on custom pricing rather than the public rate card.

No longer. As of July 21, 2026 Cerebras replaced its open free tier with a Free Trial that grants $5 in one-time credits after you create an account, with access to all Cerebras-powered models and Discord support. The docs add two conditions the pricing page leaves out: the credits are granted only after you add a verified payment method (adding one is free), and they expire 30 days after they are granted. That $5 is worth roughly 14 million input tokens on GPT-OSS-120B, and it does not renew. Past the trial, the self-serve Developer tier starts at just $10 and offers 10x higher rate limits and higher-priority processing; the Enterprise tier (contact sales) adds custom weights, dedicated queue priority, and guaranteed uptime.

Cerebras Inference delivers 1,000–2,100 tokens per second on Llama 3.1 70B-class models, compared to 40–80 tokens/second on GPU-based providers like Together AI or Fireworks AI. The speed advantage comes from the on-chip SRAM of the WSE eliminating GPU memory bandwidth bottlenecks.

As of August 2026, the public Cerebras per-token rate card lists just GPT-OSS-120B (production) and Google Deepmind Gemma 4 31B (Preview) — ZAI-GLM-4.7 was removed from the card on its disclosed August 17, 2026 deprecation date. A much wider catalog — including Llama 3.3 70B, Llama 4 Maverick and Scout, Qwen3-32B and Qwen3-235B, Qwen3-Coder, Mistral, DeepSeek, Kimi K2.x, and Z.AI's GLM 4.X/5.X families (including the former GLM-4.7) — is available through Dedicated Endpoints on reserved-capacity custom pricing. The catalog focuses on open-source models that benefit most from Cerebras's speed advantage.

Yes — it's completely free with no signup. It uses the latest publicly available Cerebras pricing (verified August 2026). Actual costs may vary with volume discounts or enterprise terms — always confirm on Cerebras's official pricing page.