AI BriefingMinimax Unveils M3 MoE Model
·2026.06.20 08:50
Key point
Minimax has open-sourced the weights of the M3 MoE model, with 428B total parameters and 23B active parameters.
Details
Minimax has open-sourced the weights of the M3 model on Hugging Face. This model adopts a MoE (Mixture of Experts) architecture, holding a total of 428B parameters, of which only 23B parameters are activated during inference.
The key features are as follows.
- Multimodal Support: It has native multimodal capabilities that can process image-text-to-text inputs.
- Efficient Architecture: Despite the small number of active parameters, it is designed to achieve performance competitive with existing 800B~2T scale models.
- Long Context: It supports a context window of 1M (1 million) tokens.
- License: The
minimax-communitylicense applies.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Xiaomi Releases Open-Weight Multimodal Model MiMo-V2.6-Pro… Higher Intelligence and Competitive Pricing Compared to Similar-Sized Models
Jevper Released: A Jev Classification Wrapper for OpenAI-Compatible Models with Automatic Logprobs Detection