Model API

The Macaron hosted model API serves the Macaron-V1 model family over an OpenAI-compatible endpoint. You call it with an API key created in the dashboard, and it exposes the same chat-completions surface that the MinT dashboard's "usage methods" panel configures for curl, Claude Code, Codex, OpenCode, and ccswitch.

This is distinct from the self-hosted MinT inference surface (/oai/api/v1 against a local MinT server with Qwen3 base models), documented under the Self-hosted tab.

Regional endpoints

Pick the base URL that matches your region:

Region	Base URL
Outside Mainland China	https://mint.macaron.im
Mainland China	https://mintcn.macaron.xin

API paths append to the base URL, e.g. {base_url}/v1/chat/completions.

The Macaron model family

Macaron-V1 is an open agent-model family built on a Mixture-of-LoRA architecture: a frozen base model plus a small set of specialist LoRA adapters, with a router selecting one specialist per user turn. The hosted lineup includes the 748B flagship Venti, the 50B Tall, the coding-specialized Coding-Venti, and the GLM-5.2 base.

For the full model family, architecture, and benchmark summary, see What is Macaron. For the hosted model list and pricing, see Models.

Get started

1. Create an API key in the dashboard (/dashboard → API Keys). Keys are sk--prefixed. See Authentication.
2. Pick a model from the hosted lineup.
3. Integrate using one of the six supported methods in Integrations.
4. Monitor usage via the dashboard stats described in Usage.

Quick example

curl https://mint.macaron.im/v1/chat/completions \
  -H "Authorization: Bearer <MACARON_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "macaron-v1-venti",
    "messages": [
      {"role": "user", "content": "Why is the sky blue?"}
    ],
    "stream": true,
    "max_completion_tokens": 1024
  }'

Replace <MACARON_API_KEY> with your dashboard key and mint.macaron.im with the Mainland-China endpoint (mintcn.macaron.xin) if applicable.