Meituan Open-Sources LongCat-2.0: 1.6 Trillion-Parameter Coding Model Trained Entirely on Chinese Chips

My take: This one left me with my jaw on the floor and I will tell you why. Meituan, best known as China's food delivery app, just open-sourced LongCat-2.0: a 1.6 trillion-parameter agentic coding model trained 100% on Chinese chips, not a single Nvidia in sight. On SWE-bench Pro (the toughest software engineering exam out there) it scores 59.5 and beats GPT-5.5 (58.6), with a 1M token context window and MIT license so you can use it commercially for anything. It is like your neighbor starting to manufacture their own cement because you stopped selling it to them, and their cement came out cheaper and better 😅. The US chip export controls that have been rolling out for years were aimed at preventing exactly this. I build with AI models every day and I see this as a signal I cannot ignore: the gap is closing, Chinese open-source models are pushing to the top, and the regulators are running to catch up. That is just my opinion.