OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Minimax M2

MiniMax-M2 is an open-source mixture-of-experts language model designed for coding, agentic tool use and general language tasks. It has 230 billion total parameters, with 10 billion active parameters.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Minimax M2 — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Consider MiniMax-M2 when building or evaluating a coding or tool-using assistant and you need an open-source model. The repository links separately to hosted API access and a MiniMax Agent experience.

Who it is for

  • Developers building coding assistants or code-repair workflows.
  • Teams integrating a language model into systems that call tools.
  • People evaluating open-source models for general language tasks.

Documented uses

  • Power coding assistance, including multi-file edits and code repair.
  • Support agent workflows that plan and use tools.
  • Provide a model for general-purpose language tasks.
Before you choose

Requirements & limits

The repository identifies the model as open source and links to an MIT license, Hugging Face, ModelScope, MiniMax API documentation and MiniMax Agent.

  • The model is a component for applications, not itself a coding IDE or agent interface.
  • The README recommends temperature 1.0, top_p 0.95 and top_k 40 for inference.
  • Applications should retain and pass back the model’s ... content in conversation history.
  • The repository links to a separate tool-calling guide for tool-enabled implementations.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Preserve MiniMax-M2 thinking content in conversation history

Use MiniMax-M2 while retaining the assistant’s interleaved thinking content across turns.

Before you start

  • A way to send MiniMax-M2 requests and pass conversation history between turns.

Steps

  1. Use the recommended inference settings: temperature 1.0, top_p 0.95, and top_k 40.

  2. When receiving an assistant response, retain any ... content in its original format.

  3. Pass that retained assistant content back in the historical messages on subsequent turns; do not strip the thinking section.

Expected outcome

Conversation history preserves the thinking content that the model’s documentation says should be retained for subsequent use.

Guide limitations

  • Removing ... from the history may negatively affect model performance.
  • The guide does not specify a request format or provide an execution example for MiniMax-M2.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
MiniMax-AI/MiniMax-M2 · Official homepage title · retained in its source language
How we review sourcesSuggest a correction