MiniMaxAI/MiniMax-M3.1-Flash
AI model · https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash
First seen by Oriel on .
AI summary
MiniMax-M3.1-Flash is a 428B-parameter native multimodal mixture-of-experts model aimed at coding, agentic workflows, computer use, and cybersecurity, with up to 1M tokens of context.
- It activates 23B parameters per token and accepts text, image, and video inputs, producing text output.
- Compared with MiniMax-M3, it uses sparse attention throughout, quantization-aware training, and DSpark speculative decoding to improve inference efficiency.
- Reported scores include 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld 2.0, and 82.2 on CyberGym.
- Local serving is recommended through SGLang, vLLM, or TokenSpeed; usable context length and throughput depend on GPU memory and serving configuration.
- Benchmark results use official weights, and third-party quantizations may differ.
#mixture of experts #multimodal models #llm inference #agentic workflows
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 537, latest observation ; +432 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 23,530, latest observation ; +23,173 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | MiniMaxAI/MiniMax-M3.1-Flash | https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash Read discussion |