MiniMax Building China’s Largest AI Model Yet—And Open-Sourcing It
MiniMax is constructing a significant artificial intelligence (AI) model in China, slated to be the biggest within the nation, featuring 2.7 trillion parameters, six times its current flagship model, according to The Information. The company intends to open-source this system in the third quarter, marking another blow to the paid frontier of AI models.
Key Takeaways:
- Model Size: MiniMax’s upcoming model boasts 2.7 trillion parameters, dwarfing its current 428 billion-parameter M3 flagship.
- Open-Sourcing: The company aims to release the model as open-source code, potentially as early as Q3 2026.
- Market Impact: This move intensifies competition within China’s AI landscape and poses a challenge to US labs’ business models, as Chinese open models are gaining traction globally.
- Commercial Importance: MiniMax, listed on the Hong Kong Stock Exchange, seeks to differentiate itself in a challenging market by showcasing its capabilities through this powerful open model.
The Landscape:
MiniMax faces strong competition from peers like DeepSeek, Zhipu, and Moonshot AI within China. Globally, Chinese open models are becoming prevalent, offering developers and enterprises cost-effective alternatives to US frontier systems. This trend has sparked concerns among American AI labs over potential revenue erosion.
Why It Matters:
A 2.7-trillion parameter open model would be a significant achievement, providing developers worldwide with a powerful, free resource and increasing pressure on the paid AI model market. Success for MiniMax could solidify its position in China and beyond, while any shortcomings could leave the open-weight crown up for grabs.