Cerebras
Cerebras-hosted models for chat, Responses and text completions.
Use models hosted on Cerebras through DeepintShield, with your Cerebras key stored on the gateway.
What works
- Chat Completions, and the Responses API through chat
- Text completions, including streaming
- List models, tool calling and SSE streaming
Limits to know: No embeddings, image generation, speech, transcription, files or batch. The upstream API is not a complete OpenAI API implementation, and reasoning.max_tokens is not accepted.
How to connect
Add your Cerebras API key in Model Hub → AI Providers, create a Virtual Key, then call cerebras/<model> through https://app.deepintshield.com/v1 with any OpenAI-compatible client.
Read the Cerebras provider docs. Requests get the same Virtual Key budgets, guardrails and logs as every other provider on the AI gateway.