What are the hardware requirements for MiniMax M3?

MiniMax M3 requires approximately 566.4 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed MiniMax M3 and how large is it?

MiniMax M3 is developed by MiniMaxAI, a model with 427b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does MiniMax M3 perform on standard benchmarks?

MiniMax M3 has the following published scores: LMSYS ELO 1445. Compare against reference models on this page or on the LLM Explorer leaderboards.

MiniMax M3 by MiniMaxAI — VRAM 566.4GB

Name: MiniMax M3
Author: MiniMaxAI

MiniMax M3 is an open-source language model by MiniMaxAI. Features: 427b LLM, VRAM: 566.4GB, License: other, LLM Explorer Score: 0.6, ELO: 1445.

Arxiv:2606.13392 Agent Coding Conversational Custom code Endpoints compatible Eval-results Image-text-to-text Minimax m3 vl Moe Multimodal Region:us Safetensors Sharded Tensorflow Video

Model Card on HF 🤗: https://huggingface.co/MiniMaxAI/MiniMax-M3

MiniMax M3 Benchmarks

LMSys ELO 1445 vs 1272 (so35) 13.6%

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

MiniMax M3 Parameters and Internals

LLM Name	MiniMax M3
Repository 🤗	https://huggingface.co/MiniMaxAI/MiniMax-M3
Model Size	427b
Required VRAM	566.4 GB
Updated	2026-08-08
Maintainer	MiniMaxAI
Model Type	minimax_m3_vl
Model Files	5.6 GB: 1-of-59 10.9 GB: 2-of-59 16.1 GB: 3-of-59 16.1 GB: 4-of-59 16.1 GB: 5-of-59 16.1 GB: 6-of-59 16.1 GB: 7-of-59 16.1 GB: 8-of-59 16.1 GB: 9-of-59 16.1 GB: 10-of-59 16.1 GB: 11-of-59 16.1 GB: 12-of-59 16.1 GB: 13-of-59 16.1 GB: 14-of-59 16.1 GB: 15-of-59 16.1 GB: 16-of-59 16.1 GB: 17-of-59 16.1 GB: 18-of-59 16.1 GB: 19-of-59 16.1 GB: 20-of-59 16.1 GB: 21-of-59 16.1 GB: 22-of-59 16.1 GB: 23-of-59 16.1 GB: 24-of-59 5.2 GB: 25-of-59 15.3 GB: 26-of-59 14.8 GB: 27-of-59 14.8 GB: 28-of-59 14.8 GB: 29-of-59 14.8 GB: 30-of-59 14.8 GB: 31-of-59 14.8 GB: 32-of-59 14.8 GB: 33-of-59 14.8 GB: 34-of-59 14.8 GB: 35-of-59 14.8 GB: 36-of-59 13.6 GB: 37-of-59 13.6 GB: 38-of-59
Model Architecture	MiniMaxM3SparseForConditionalGeneration
License	other
Transformers Version	4.52.4
Torch Data Type	bfloat16

Rank the MiniMax M3 capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Quantized Models of the MiniMax M3

Model	Likes	Downloads	VRAM
MiniMax M3 4bit	6	2304	173 GB

Best Alternatives to MiniMax M3

Best Alternatives	Context / RAM	Downloads	Likes
MiniMax M3	0K / 593.6 GB	187	8
MiniMax M3 Uncensored	0K / 580 GB	884	22