Skip to content

Encrypted inference

API

Explore the OpenAI-compatible endpoint and private transport before creating keys, adding credits, and monitoring usage.

Sign in to unlock the full API console
OverviewReference

Models

Use a base model name or add tee-. GET /v1/models returns only executable paid aliases.

  • gpt-oss-120bFast reasoning for agents, code, and everyday work.Text · reasoning
  • kimi-k2.6A heavier model for demanding tasks and image input.Text · images · reasoning

Endpoints

The paid API currently supports standard and explicit TEE Chat Completions. Browser-only E2EE features are intentionally separate.

OpenAPI 3.1
  • Chat CompletionsStandard JSON and explicit TEE aliases; streaming, tools, structured outputs, and vision.Supported
  • CompletionsNot exposed until a paid, verifiable transport is available.Not yet
  • Anthropic MessagesNot exposed until a paid, verifiable transport is available.Not yet
  • Document parsingAvailable in signed-in browser chat, not the paid API.Not yet
  • Responses APIThe private backend does not expose this route yet.Not yet
  • Image, video, and text-to-speech generationNo confidential generation backend is available yet.Not yet

One endpoint, two modes

The base URL and API key never change. Pick the paid privacy contract in the model name.

  1. model

    Ordinary OpenAI JSON. Garlic encrypts the request into its confidential inference backend.

  2. tee-model

    The same JSON shape with an explicit confidential-compute contract.