Models like GLM-5.2 and Kimi K3 are narrowing the gap with the US’s best. And according to Guohao Li, founder of open-source AI community CAMEL-AI, the secret weapon might be something unexpected: interns.
Li points out that Chinese undergraduate and graduate interns are often directly involved in training models. They get hands-on with the details and are more willing to share what they learn publicly.
In contrast, he sees US frontier labs as much more closed. Training methods and compute access are concentrated in a small number of researchers, and even top PhD students often struggle to get internship spots.
For Li, China’s advantage isn’t just about open-weight models — it’s about a more open talent pipeline and the free flow of experience.
https://twitter.com/guohao_li/status/2078538012288221490?s=46