跳转到主要内容

MiniMax M1 40k

MiniMaxMiniMax开源权重MIT · 商用许可

描述

MiniMax-M1 is an open-source, large-scale reasoning model that uses a hybrid-attention architecture for efficient long-context processing. It supports up to a 1 million token context window and 80,000-token reasoning output, matching Gemini 2.5 Pro’s scale while being highly cost-effective. Its Lightning Attention mechanism reduces compute requirements to about 30% of DeepSeek R1’s, and a new reinforcement learning algorithm, CISPO, doubles convergence speed compared to other RL methods. Trained on 512 H800s over three weeks, M1 achieves near state-of-the-art results across software engineering, long-context, and tool-use benchmarks, outperforming most open models and rivaling top closed systems.

发布日期
2025-06-17
参数规模
456.0B
上下文长度
—
支持模态
—

能力雷达图

60
general
60
coding
70
reasoning
60
science估算
60
agents
10
multimodal

缺少专门科学评测时,Science 由 LLM Stats 科学得分或推理能力估算。

排行榜排名

暂无排名数据

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Chat

TAU-bench Retail67.8%自报
Multi-Challenge44.7%自报

Factuality

SimpleQA17.9%自报

Language

MMLU-Pro80.6%自报

Long Context

OpenAI-MRCR: 2 needle 128k76.1%自报
LongBench v261.0%自报
OpenAI-MRCR: 2 needle 1M58.6%自报

Math

MATH-50096.0%自报
AIME 202483.3%自报
AIME 202574.6%自报

Reasoning

ZebraLogic80.1%自报
GPQANYU + Cohere + Anthropic (2023)69.2%自报
LiveCodeBench62.3%自报
TAU-bench Airline60.0%自报
SWE-Bench Verified55.6%自报
Humanity's Last Exam7.2%自报

AA 评测指数

(Artificial Analysis)

暂无 AA 评测数据

LLM Stats 分类评分

(LLM Stats (zeroeval))
Language
80
Legal
80
Finance
80
Healthcare
80
Long Context
70
Math
70
Physics
70
Biology
70
Chemistry
70
Chat
60
Reasoning
60
Structured Output
60
Frontend Development
60
General
60
Code
60
Communication
60
Tool Calling
60
Factuality
20
Vision
10

定价

暂无定价数据

速度

暂无速度数据

供应商价格排行

暂无提供商数据

外部链接