The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence
The MiniMax-M2 series introduces a new family of Mixture-of-Experts language models. These models leverage mini activations to maximize real-world intelligence, with the flagship model featuring 229.9 billion parameters. The design focuses on agent-driven data pipelines for effective deployment and performance.
- ▪The MiniMax-M2 series consists of advanced language models that utilize mini activations.
- ▪The flagship model, M2, has a total of 229.9 billion parameters with only 9.8 billion activated per token.
- ▪The models are designed for agentic deployment, emphasizing large-scale data pipelines.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.26494 (cs) [Submitted on 26 May 2026] Title:The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence Authors:MiniMax: Aili Chen, Aonian Li, Baichuan Zhou, Bangwei Gong, Binyang Jiang, Boji Dan, Changqing Yu, Chao Wang, Cheng Ma, Cheng Zhong, Cheng Zhu, Chengjun Xiao, Chengyi Yang, Chengyu Du, Chenyang Zhang, Chi Zhang, Chuangyi Huang, Chunhao Zhang, Chunhui Du, Chunyu Zhao, Congchao Guo, Da Chen, Deming Ding, Dianjun Sun, Dongyu Zhang, Enhui Yang, Fei Yu, Guang Zheng, Guodong Zheng, Guohong Li, Haichao Zhu, Haigang Zhou, Haimo Zhang, Han Ding, Hao Zhang, Haohai Sun, Haolin Lyu, Haonan Lu, Haoyu Wang, Huajie Shi, Huiyang Li, Jiacheng Chen, Jian Zhang, Jiaqi Zhuang, Jiaren Cai, Jiaxin Pan, Jiayao Li, Jiayuan Song,…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
