MiniMax

Use this provider to route to MiniMax's hosted models. Aurora talks to MiniMax through the OpenAI-compatible endpoint.

Activate

MINIMAX_API_KEY=...

Restart Aurora. The MiniMax provider appears at /v1/models with MiniMax's chat, vision, and speech model catalog.

Default models

Override with MINIMAX_MODELS=MiniMax-m1,MiniMax-vl-01.

Multiple instances

MINIMAX_PROD_API_KEY=...
MINIMAX_STAGING_API_KEY=...

Aurora exposes them as MiniMax-prod and MiniMax-staging.

Custom base URL

MINIMAX_BASE_URL=https://api.minimax.io/v1

Gotchas

• Speech / TTS — MiniMax's TTS endpoint is not on the OpenAI-compatible surface. Use provider passthrough at /p/MiniMax/* to call the native API.
• Reasoning models — MiniMax reasoning models return reasoning content blocks. Aurora preserves them.
• Vision — MiniMax-vl-01 accepts image inputs in the OpenAI image_url content shape. Aurora forwards them.
• Function calling — full support, no quirks.
• Prompt caching — not supported by MiniMax. Use Aurora's semantic cache if you need replay.

Sample request

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AURORA_MASTER_KEY" \
  -d '{
    "model": "MiniMax/MiniMax-m1",
    "messages": [{"role": "user", "content": "Hello from Aurora"}]
  }'

Related

• Providers Overview
• Passthrough API — for native MiniMax features