Key facts
- Chen co-developed the Attention Residuals architecture for Kimi K3.
- Kimi K3 is a model with 2.8 trillion total parameters.
- Chen plans to relocate to the San Francisco Bay Area.
Nathan Chen, a 17-year-old AI researcher, announced on Thursday that he has left Moonshot AI, where he spent less than a year. During his tenure, Chen co-developed the Attention Residuals (AttnRes) architecture, which has been integrated into the Kimi K3 model, a large language model with 2.8 trillion parameters. Chen's departure has fueled speculation about his next steps, particularly amid intense competition for AI talent between Silicon Valley and China. He indicated plans to relocate to the San Francisco Bay Area. Chen's journey into AI research began with a hackathon presentation, leading to internships and eventually a role at Moonshot AI.