Head of DevRel @MiniMax_AI. Building @MiniMaxAgent and @Hailuo_AI. Make more model be open
SH/SF
Joined May 2025
- Tweets605
- Following359
- Followers12.1K
- Likes3K
Enjoy the weight: huggingface.co/MiniMaxAI/Min…
MiniMax M3.1 Flash weights are now public. 428B parameters total, 23B activated per token.
Agentic performance: A generational leap in agentic tasks, coding, and cybersecurity — 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam, and 82.2 on CyberGym.
Efficient by design: MiniMax Sparse Attention and DSpark speculative decoding improve inference efficiency.
Hugging Face:
huggingface.co/MiniMaxAI/Min…
ModelScope:
modelscope.cn/models/MiniMax…
API: platform.minimax.io
M Plan: minimax.io/m-planMiniMax
MCode for desktop: agent.minimax.io/download
I’ll be in the Bay Area next week!
Open for coffee chats with folks/teams working on model evaluation, RL environment construction.
Feel free to DM me if you’re around. Please RT and share this with anyone who might be interested.
We’re gearing up for something big.
#AI #LLM #RL #ModelEvaluation #Inference
RyanLee retweeted
We were working on our RSI Arena report when we saw that MiniMax M3.1 Flash launched today. Had to share this one 🐰: #2 with a remarkably lower cost compared to GPT-6 Astra.
It really surprised us with how well it could train another AI. It ran experiments, trained a model, then kept improving it with human feedback.
We’re wrapping up the technical report and will release all training artifacts and model checkpoints soon! Stay tuned!
