Is MiniMax’s new AI model surpassing Google DeepMind?

The Chinese start-up MiniMax has launched the M2, a groundbreaking AI model that has achieved a top ranking on a global leaderboard for open models. The M2 has earned an impressive score on Artificial Analysis’s intelligence index, placing it among the world’s top five models. It has outperformed Google DeepMind’s Gemini 2.5 Pro, released in June, and is positioned just below leading models from US companies like OpenAI, Anthropic, and xAI. With a massive 200 billion parameters, the M2 utilizes a Mixture-of-Experts architecture, enhancing the efficiency of Chinese AI models. Notably, it activates only 10 billion parameters per forward pass, a trend aimed at improving efficiency. This approach allows the model to function effectively on a large scale, unlike other models such as DeepSeek’s V3.2 and Moonshot AI’s Kimi K2, which use more active parameters. Florian Brand, a PhD student and open models expert, praised the significant advancements from MiniMax’s previous model.