Skip to main content
7 min readPublished: 2026-08-10Updated: 2026-08-10

MiniMax China API Key Guide: M2.7, Pricing and First Call

Use the official MiniMax China key page, compare current M2.7 pay-as-you-go prices, and make a first request with the OpenAI- or Anthropic-compatible API.

By Heizi· Founder & Editor· Published: 2026-08-10Hands-on tested

Official MiniMax China API key page

Sign in to https://platform.minimaxi.com and open Account Management → Interface Keys. The direct official key URL is https://platform.minimaxi.com/user-center/basic-information/interface-key. Create a separate key for each application and keep it on the server.

Current M2 text models

MiniMax's current API overview lists M2.7, M2.7-highspeed, M2.5, M2.5-highspeed, M2.1, M2.1-highspeed, and M2. All have a documented 204,800-token input-plus-output window. M2.7 is described at about 60 output tokens per second, while M2.7-highspeed is about 100 TPS.

Verified pay-as-you-go text pricing

Prices below are CNY per million tokens from MiniMax's official pay-as-you-go page, checked August 10, 2026. Audio, image, video and music use separate units and price tables.

  • MiniMax-M2.7: input ¥2.1; output ¥8.4; cache read ¥0.42; cache write ¥2.625
  • MiniMax-M2.7-highspeed: input ¥4.2; output ¥16.8; cache read ¥0.42; cache write ¥2.625
  • MiniMax-M2.5: input ¥2.1; output ¥8.4; cache read ¥0.21; cache write ¥2.625
  • MiniMax-M2.5-highspeed: input ¥4.2; output ¥16.8; cache read ¥0.21; cache write ¥2.625
  • M2-her: input ¥2.1; output ¥8.4; no cache price listed

First call with the OpenAI-compatible API

MiniMax officially supports the OpenAI SDK at https://api.minimaxi.com/v1. The response content can include thinking tags, so preserve the full assistant message in multi-turn conversations.

python
from openai import OpenAI
import os

client = OpenAI(
    api_key=os.environ["MINIMAX_API_KEY"],
    base_url="https://api.minimaxi.com/v1",
)
response = client.chat.completions.create(
    model="MiniMax-M2.7",
    messages=[{"role": "user", "content": "Reply with: KEY_WORKS"}],
)
print(response.choices[0].message.content)

MiniMax M2.7 with the OpenAI Python SDK

Anthropic-compatible endpoint

MiniMax also provides an Anthropic-compatible endpoint at https://api.minimaxi.com/anthropic and recommends this integration for current text models. It supports text, tool use, tool results and thinking blocks; the compatibility page says image and document inputs are not currently supported.

FAQ

What is the MiniMax China API base URL?

Use https://api.minimaxi.com/v1 for the OpenAI SDK or https://api.minimaxi.com/anthropic for the Anthropic SDK.

How large is the M2.7 context window?

MiniMax documents a 204,800-token combined input and output window.

Is M2.7-highspeed a different-quality model?

MiniMax describes it as the same capability with faster output, approximately 100 TPS instead of approximately 60 TPS, and prices it at twice the standard input/output rate.

Related Providers

Sources