MiniMax releases MiniMax-M3.1-Flash, a 428B multimodal MoE
Original title: MiniMaxAI/MiniMax-M3.1-Flash — new model trending #30 on Hugging Face
About 23B parameters activate per token; inputs include text, image, and video, with a 1 million token context window.
MiniMax published MiniMax-M3.1-Flash, a native multimodal mixture-of-experts model for coding, agentic workflows, computer use, and cybersecurity. It builds on MiniMax-M3 with sparse attention throughout, quantization-aware training, and DSpark speculative decoding. The model card lists vendor scores, including 76.2 on BankerToolBench and 82.2 on CyberGym. It is trending on Hugging Face and is offered through MiniMax's API.
Unconfirmed: Benchmark figures are MiniMax's own card scores and are not independently verified here.
- Total parameters are about 428B, with 23B activated per token.
- The card lists 62.8 on Automation-Bench v1.0.6 and 61.7 on OSWorld V2.
- Other listed scores are 73.7% on DeepSWE v1.1 and 29.6 on Agents' Last Exam.
- A DSpark draft head is included for speculative decoding.
- Vendors
- MiniMax
- AI models
- MiniMax-M3.1-Flash
From the original article
4 min read at huggingface.coimage-text-to-text · 357 downloads · 156 likes · license other
By MiniMaxAI · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.