MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention

We introduce MiniMax-M1, the world’s first open-weight, large-scale hybrid-attention reasoning model. MiniMax-M1 is powered by a hybrid Mixture-of-Experts (MoE) architecture combined with a lightning attention mechanism. The model is developed based on our previous MiniMax-Text-01 model (MiniMax et al., 2025), which contains a total of 456 billion parameters with 45.9 billion parameters activated per token. The M1 model natively supports a context length of 1 million tokens, 8x the context size of DeepSeek R1. Furthermore, the lightning attention mechanism in MiniMax-M1 enables efficient scaling of test-time compute – For example, compared to DeepSeek R1, M1 consumes 25% of the FLOPs at a generation length of 100K tokens. These properties make M1 particularly suitable for complex tasks that require processing long inputs and thinking extensively. MiniMax-M1 is trained using large-scale reinforcement learning (RL) on diverse problems ranging from traditional mathematical reasoning to sandbox-based, real-world software engineering environments. In addition to the inherent efficiency advantage of lightning attention for RL training, we propose CISPO, a novel RL algorithm to further enhance RL efficiency.

Introduction. Large reasoning models (LRMs), such as OpenAI o1 (OpenAI, 2024a) and DeepSeek-R1 (DeepSeek-AI et al., 2025), have demonstrated remarkable success by extending the length of reasoning through large-scale reinforcement learning (RL). In recent months, both the open-source community and commercial organizations have followed this trend, achieving significant advances on complex tasks such as Olympiad mathematics competitions and competitive programming (Anthropic, 2025; Google DeepMind, 2025; Hu et al., 2025; Kimi Team, 2025; Seed et al., 2025; Yu et al., 2025; Zeng et al., 2025). The success of LRMs has been primarily attributed to a new scaling dimension of test-time compute—As more FLOPs are dedicated to extended reasoning processes during generation, model performance shows consistent improvement, particularly for complex real-world applications (Jimenez et al., 2024; OpenAI, 2025).

Discussion / Conclusion. In this work, we introduce and release MiniMax-M1, the world’s first open-weight, large-scale reasoning model featuring a lightning attention mechanism. This efficient attention design enables MiniMax-M1 to natively support inputs of up to 1M tokens and generation lengths of 80K tokens—both significantly exceeding capabilities of other open-weight models. These capabilities render MiniMax- M1 uniquely suited for complex, realistic scenarios requiring long context and extended reasoning, properties empirically validated by its strong performance on software engineering, agentic tool use, and long-context understanding benchmarks. Beyond the inherent efficiency advantages of lightning attention for RL training, this work contributes a novel RL algorithm, CISPO, to accelerate training. Combining architectural advantages with CISPO, we efficiently trained MiniMax-M1, with complete RL training completed in three weeks using 512 H800 GPUs. Across comprehensive evaluations, MiniMax-M1 ranks among the world’s best open-weight models alongside DeepSeek-R1 and Qwen3- 235B.