MiniMax-Text-01 — Demo Space

Temperature

0 (precise)0.71.5 (creative)

Max tokens

Stream

Demo mode simulates responses. Download with your MINIMAX_API_KEY to go live.

Live endpoint

https://api.minimax.chat/v1/text/chatcompletion_v2

curl https://api.minimax.chat/v1/text/chatcompletion_v2 \
  -H "Authorization: Bearer $MINIMAX_API_KEY" \
  -d '{
    "model": "MiniMax-Text-01",
    "messages": [{"role":"user","content":"Hello, 4M context!"}],
    "temperature": 0.7,
    "max_tokens": 512
  }'

Example response

{
  "id": "5cff1c5e-d09f-4a7d-9f8c-2f2d6b4c3a1e",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "Hello! I can help with that. 🚀"
    },
    "finish_reason": "stop"
  }],
  "usage": { "total_tokens": 18 }
}

Settings hint

Set the MINIMAX_API_KEY secret in your Space settings and toggle useLiveApi = true at the top of the script to hit the real endpoint. Note: browsers may hit CORS — use the Gradio/curl path for production deployments.

Context window

4M tokens

≈ several full-length novels in a single prompt.

Parameters

456B

Hybrid Transformer + linear attention for efficiency.

Highlights

4M-token context window — among the largest in open models.
Achieves GPT-4o / Claude 3.5 Sonnet level performance on many benchmarks.
Uses lightning attention for long-sequence efficiency.
Released on Hugging Face under the MiniMaxAI/MiniMax-Text-01 repo.
MIT-ish commercial-friendly open model license.