Cerebras Inference API

aid: cerebras:cerebras-inference-api
name: Cerebras Inference API
description: The Cerebras Inference API exposes ultra-low-latency inference for open-weight large language
  models including Llama 3.1, Llama 4, Qwen, and other frontier open models. The API is OpenAI-compatible
  at the chat completions surface, supports streaming, and is consumed via first-party Python and Node.js
  SDKs as well as raw HTTP. Dedicated and on-prem deployments are available for production workloads.
humanURL: https://inference-docs.cerebras.ai
baseURL: https://api.cerebras.ai/v1
tags:
- Inference
- LLM
- Chat Completions
- OpenAI Compatible
- Streaming
- REST
properties:
- type: Documentation
  url: https://inference-docs.cerebras.ai
- type: GettingStarted
  url: https://inference-docs.cerebras.ai/quickstart
- type: SDK
  url: https://github.com/Cerebras/cerebras-cloud-sdk-python
- type: SDK
  url: https://github.com/Cerebras/cerebras-cloud-sdk-node
- type: Cookbook
  url: https://github.com/Cerebras/Cerebras-Inference-Cookbook
- type: VSCodeExtension
  url: https://github.com/Cerebras/vscode-cerebras-chat
- type: MCP
  url: https://github.com/Cerebras/cerebras-code-mcp
features:
- name: OpenAI-Compatible Chat Completions
  description: Drop-in compatibility with OpenAI client libraries for fast migration of existing applications.
- name: Ultra-Fast Token Generation
  description: WSE-3 wafer-scale silicon delivers token-per-second throughput marketed as up to 15x faster
    than GPU inference.
- name: Open-Weight Model Catalog
  description: Hosted access to Llama, Qwen, DeepSeek, and other curated open-source models with no infrastructure
    setup.
- name: Streaming Responses
  description: Server-sent event streaming for chat completions enabling real-time agent and voice UX.
- name: Dedicated Endpoints
  description: Private capacity and custom model hosting via dedicated endpoint tier for production workloads.
- name: First-Party SDKs
  description: Official Python and TypeScript/Node SDKs with typed model and parameter support.
- name: On-Premises Deployment
  description: CS-2 and CS-3 systems for private data center and sovereign AI deployments.
useCases:
- name: Real-Time Voice and Agent Applications
  description: Power voice agents, copilots, and tool-calling agents that need sub-second time-to-first-token.
- name: Coding Copilots
  description: Drive code generation and review assistants with fast inference on open-weight coding models.
- name: Reasoning and Research Workloads
  description: Run long-context reasoning loops and chain-of-thought workflows economically at high throughput.
- name: Enterprise Inference Migration
  description: Move existing OpenAI-based workloads to Cerebras with minimal code change for cost and
    latency wins.
- name: Healthcare and Life Sciences
  description: Used by partners including GSK and Mayo Clinic for biomedical and clinical AI workloads.
integrations:
- name: OpenAI SDK
- name: LangChain
- name: LlamaIndex
- name: Vercel AI SDK
- name: AWS
- name: Hugging Face
- name: VS Code
- name: Model Context Protocol
authentication:
- type: API Key
  description: Requests authenticate via Bearer token using a CEREBRAS_API_KEY provisioned from the Cerebras
    Cloud dashboard.