# Cerebras > The fastest public inference host, running models on a wafer-scale chip. Canonical: https://www.subconscious.dev/providers/cerebras · By The Subconscious Team · Updated September 30, 2026 - Founded: 2015 - Example models: GPT-OSS 120B, Gemma 4 31B - Website: https://www.cerebras.ai ## Overview Cerebras builds a single chip the size of a silicon wafer, and its inference cloud is the fastest public host on the models it serves. Its developer table lists GPT-OSS 120B near 3,000 tokens per second, about six times Groq on the same weights, at $0.35 in and $0.75 out per million. The public shared catalog is thin, just GPT-OSS 120B and Gemma 4 31B as of August 2026. More model families live on dedicated endpoints and through partners like OpenRouter, Hugging Face, Vercel and AWS Marketplace. OpenAI is the anchor customer. A January 2026 deal worth over $10B has OpenAI renting roughly 750 MW of Cerebras capacity through 2028, and on August 13, 2026 OpenAI previewed an Ultrafast tier of GPT-5.6 Sol running on Cerebras at up to 750 output tokens per second. Cerebras now trades on Nasdaq as CBRS. In late September it announced CS-4 hardware and a Gimlet Labs partnership that pairs its wafers with GPUs in a disaggregated cloud targeting 3,000 tokens per second. ## Upsides - Fastest published tokens per second of any public inference host. - The only wafer-scale path to a closed frontier model, via OpenAI's Ultrafast preview. ## Core use cases - Voice, live code autocomplete and streaming UIs where generation is the wait. - Agent steps that emit long outputs on open weights. ## Downsides - Tiny self-serve catalog, so most models mean a sales conversation. - Costs more per token than Groq on shared models, and the speed does little when an agent mostly waits on tools or hidden reasoning. ## At a glance | Attribute | Value | |---|---| | Model access | Open weights | | Flagship models | GPT-OSS 120B, Gemma 4 31B | | Speed | ~3,000 tok/s on GPT-OSS 120B | | Price | $0.35 in, $0.75 out (GPT-OSS 120B) | | Customization | - | | Deployment | Shared API, dedicated, partners | | Long context | - | ## FAQ ### What is Cerebras? Cerebras builds a single chip the size of a silicon wafer, and its inference cloud is the fastest public host on the models it serves. Its developer table lists GPT-OSS 120B near 3,000 tokens per second, about six times Groq on the same weights, at $0.35 in and $0.75 out per million. The public shared catalog is thin, just GPT-OSS 120B and Gemma 4 31B as of August 2026. More model families live on dedicated endpoints and through partners like OpenRouter, Hugging Face, Vercel and AWS Marketplace. ### What is Cerebras best for? Voice, live code autocomplete and streaming UIs where generation is the wait; Agent steps that emit long outputs on open weights. ### How much does Cerebras cost? Cerebras pricing at a glance: $0.35 in, $0.75 out (GPT-OSS 120B). Rates change often, so check Cerebras's pricing page before committing. ### What are the downsides of Cerebras? Tiny self-serve catalog, so most models mean a sales conversation; Costs more per token than Groq on shared models, and the speed does little when an agent mostly waits on tools or hidden reasoning. ### What are the best alternatives to Cerebras? Common alternatives include Subconscious, OpenAI, Anthropic, Google Vertex AI, Amazon Bedrock. Each has a head-to-head comparison with Cerebras on this site. ## Comparisons - [Subconscious vs Cerebras](https://www.subconscious.dev/compare/subconscious-vs-cerebras.md) - [OpenAI vs Cerebras](https://www.subconscious.dev/compare/openai-vs-cerebras.md) - [Anthropic vs Cerebras](https://www.subconscious.dev/compare/anthropic-vs-cerebras.md) - [Google Vertex AI vs Cerebras](https://www.subconscious.dev/compare/google-vertex-vs-cerebras.md) - [Amazon Bedrock vs Cerebras](https://www.subconscious.dev/compare/aws-bedrock-vs-cerebras.md) - [Together AI vs Cerebras](https://www.subconscious.dev/compare/together-ai-vs-cerebras.md) - [Fireworks AI vs Cerebras](https://www.subconscious.dev/compare/fireworks-vs-cerebras.md) - [Baseten vs Cerebras](https://www.subconscious.dev/compare/baseten-vs-cerebras.md) - [Groq vs Cerebras](https://www.subconscious.dev/compare/groq-vs-cerebras.md) - [Cerebras vs DeepInfra](https://www.subconscious.dev/compare/cerebras-vs-deepinfra.md) - [Cerebras vs Modal](https://www.subconscious.dev/compare/cerebras-vs-modal.md) - [Cerebras vs xAI](https://www.subconscious.dev/compare/cerebras-vs-xai.md) - [Cerebras vs DeepSeek](https://www.subconscious.dev/compare/cerebras-vs-deepseek.md) - [Cerebras vs Moonshot AI](https://www.subconscious.dev/compare/cerebras-vs-moonshot-ai.md) - [Cerebras vs Z.ai](https://www.subconscious.dev/compare/cerebras-vs-z-ai.md) - [Cerebras vs Alibaba Cloud](https://www.subconscious.dev/compare/cerebras-vs-alibaba-cloud.md) - [Cerebras vs Meta](https://www.subconscious.dev/compare/cerebras-vs-meta.md) - [Cerebras vs SambaNova](https://www.subconscious.dev/compare/cerebras-vs-sambanova.md) - [Cerebras vs Nebius](https://www.subconscious.dev/compare/cerebras-vs-nebius.md) - [Cerebras vs fal](https://www.subconscious.dev/compare/cerebras-vs-fal.md) - [Cerebras vs Novita AI](https://www.subconscious.dev/compare/cerebras-vs-novita-ai.md) - [Cerebras vs Parasail](https://www.subconscious.dev/compare/cerebras-vs-parasail.md) - [Cerebras vs Inference.net](https://www.subconscious.dev/compare/cerebras-vs-inference-net.md) - [Cerebras vs GMI Cloud](https://www.subconscious.dev/compare/cerebras-vs-gmi-cloud.md) - [Cerebras vs Sail Research](https://www.subconscious.dev/compare/cerebras-vs-sail-research.md) - [Cerebras vs Morph](https://www.subconscious.dev/compare/cerebras-vs-morph.md) - [Cerebras vs Relace](https://www.subconscious.dev/compare/cerebras-vs-relace.md) - [Cerebras vs TypeSafe AI](https://www.subconscious.dev/compare/cerebras-vs-typesafe-ai.md) - [Cerebras vs StepFun](https://www.subconscious.dev/compare/cerebras-vs-stepfun.md) - [Cerebras vs Runware](https://www.subconscious.dev/compare/cerebras-vs-runware.md) - [Cerebras vs StreamLake](https://www.subconscious.dev/compare/cerebras-vs-streamlake.md) - [Cerebras vs Wafer](https://www.subconscious.dev/compare/cerebras-vs-wafer.md) - [Cerebras vs RunInfra](https://www.subconscious.dev/compare/cerebras-vs-runinfra.md) - [Cerebras vs Particle.AI](https://www.subconscious.dev/compare/cerebras-vs-particle-ai.md) ## Sources - [Cerebras vs Groq, BenchLM](https://benchlm.ai/blog/posts/cerebras-won-speed-not-the-shortlist) - [Cerebras OpenAI deal, The Next Platform](https://www.nextplatform.com/ai/2026/01/15/cerebras-inks-transformative-10-billion-inference-deal-with-openai/4092155) - [Cerebras and Gimlet Labs](https://www.cerebras.ai/press-release/gimlet-labs-adds-cerebras-to-deliver-ultrafast-ai-inference-through-gimlet-cloud-deployment) Pricing and model lineups change often; figures are a snapshot.