What the New Kimi K3 Model Really Means for the U.S.-China AI Race

Thursday’s release of a surprisingly sophisticated new open-source AI model from Beijing-based startup Moonshot has ignited debate over whether China’s developers are erasing the gap with the dominant models from America’s AI firms.

There is little question that Moonshot’s Kimi K3 is impressive. The largest open-source model in the world, with 2.8 trillion parameters, it performed impressively on some AI benchmarks, topping a key coding leaderboard from Arena, for instance—ahead of both OpenAI‘s GPT-5.6 and Anthropic’s most powerful model, Claude Fable 5. 

Anastasios Angelopoulos, co-founder and CEO of Arena, said in an interview on The Information's TITV that Kimi K3 challenges the perception that Chinese open-source models are advancing simply by learning from top U.S. models, a process known as distillation. 

“This is the first time we're really seeing the narrative that breaks that mental model, that says that the Chinese labs might actually just be really good at developing models and not just distilling American intelligence,” Angelopoulos said.