MiniMax-M3 API: the 1M-context flagship model

MiniMax's newest flagship text model with a 1,048,576-token context, tiered pricing and cache hits from $0.0627 per million tokens.

About MiniMax-M3

MiniMax-M3 is MiniMax's 2026 flagship text model with a 1,048,576-token (1M) context window, built for whole-repository code analysis, long-document understanding, multi-step agent workloads and long conversations. It posts strong results on coding benchmarks such as SWE-Bench while keeping per-token cost low.

MiniMax-M3 API pricing (USD per 1M tokens)

The tier is chosen from the total input length of a request, cached tokens included. Tokens served from the prompt cache are billed at the cache price only, so multi-turn conversations with a long system prompt or document get markedly cheaper.

Quick start

curl

curl https://yiduochan.com/v1/chat/completions \
  -H "Authorization: Bearer $YIDUOCHAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-M3",
    "messages": [{"role": "user", "content": "Introduce the MiniMax models in one sentence"}]
  }'

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(base_url="https://yiduochan.com/v1", api_key="YIDUOCHAN_API_KEY")
resp = client.chat.completions.create(
    model="MiniMax-M3",
    messages=[{"role": "user", "content": "Introduce the MiniMax models in one sentence"}],
)
print(resp.choices[0].message.content)

Use it in Claude Code and other Anthropic-compatible tools

The same key also works against our Anthropic Messages compatible endpoint /v1/messages, so MiniMax models plug straight into Claude Code or the Anthropic mode of Cline. For Claude Code, export three variables:

export ANTHROPIC_BASE_URL=https://yiduochan.com
export ANTHROPIC_AUTH_TOKEN=YIDUOCHAN_API_KEY
export ANTHROPIC_MODEL=MiniMax-M3

Where it shines

• Long-document Q&A and summarization: load a whole manual, contract or paper collection without chunking and retrieval.
• Codebase analysis and refactoring: feed the relevant files of an entire repository for cross-file bug hunting and change proposals.
• Agents and tool use: function calling plus a huge context keeps the full task state in the conversation.
• Multilingual writing and translation: balanced Chinese and English quality for content pipelines.

MiniMax-M3 or MiniMax-M2.7?

Both cost the same in the ≤512K tier. Choose MiniMax-M3 when you need the 1M context or stronger reasoning; for everyday chat, support bots and bulk generation where latency matters, MiniMax-M2.7 and M2.7-highspeed are the economical pick.

FAQ

Sign up and get an API key See full model pricing