Bỏ qua đến nội dung chính
Back to home
AI 1 min read

MiniMax-M2: A 230-Billion-Parameter AI that Only Activates 4% of Its Power

MiniMax has launched the M2 MoE model series with 229.9 billion parameters, optimized for agents and capable of self-debugging its own source code.

Tier 2 · sources 99% confidence Reviewed
Sources arxiv.org

Today, MiniMax introduced its Mixture-of-Experts (MoE) language model series named MiniMax-M2, specifically designed for AI agent tasks and self-evolution capabilities.

Developments

The flagship M2 model boasts a total of 229.9 billion parameters, but thanks to the MoE architecture, it only activates 9.8 billion parameters per token (~4.2%). This allows the model to maintain high performance while saving computational resources. The breakthrough lies in the M2.7 checkpoint, where the model begins to exhibit self-evolution capabilities: it can automatically debug its training runs and modify its own scaffold.

Why It Matters

The ability to 'self-debug' and 'self-modify structure' is a significant step toward AGI. For AI engineers in Vietnam, the release of the M2 series—optimized for agentic deployment—opens up opportunities to build more complex autonomous systems, ranging from coding to office assistance, without being burdened by the massive operational costs of traditional dense models.

However, this 'self-evolution' capability also raises new questions regarding AI safety and model alignment as they begin to alter their own source code.