The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence

The MiniMax-M2 series is a family of Mixture-of-Experts language models built around the principle that mini activations can unleash maximum real-world intelligence: the flagship M2 has 229.9B total parameters with only 9.8B activated per token. Designed end-to-end for agentic deployment, the series rests on agent-driven data pipelines producing large-scale verifiable trajectories grounded in executable workspaces, and Forge, a scalable agent-native RL system for long-horizon agent trajectories.

M2 achieved 69% Pass@1 on SWE-bench Verified and ranked #2 on MultiSWE and TerminalBench. During my internship on the base model team, I contributed to code-agent post-training across M1, M2, M2.1, and M2.5 — including large-scale SWE data synthesis (36K verifiable tasks from 5K+ sandbox environments), a rubric-based evaluation benchmark built from live user feedback, and CodeMirror ToolScaling for M2.5.

Paper · Release notes

Recommended citation:

@article{minimax2026m2,
  title={The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence},
  author={MiniMax},
  journal={arXiv preprint arXiv:2605.26494},
  year={2026}
}