AI · 2h ago
China's Moonshot AI drops 2.8T-parameter Kimi K3, beating Claude on coding benchmark
Moonshot AI released Kimi K3, a 2.8 trillion parameter open-weight model, claiming it outperforms Anthropic's Claude on the Frontend Code Arena benchmark. The model is the largest open-weight AI ever, developed despite US export restrictions on advanced chips. Moonshot says Kimi K3 is the first open 3T-class system.
Meridian48 take
The claim of beating Claude is notable, but benchmarks can be narrow; the real story is China's ability to train frontier models under compute sanctions.
Read the full reporting
China's 2.8-trillion-parameter Kimi K3 beats Claude Fable 5 in Frontend Code Arena benchmark— Moonshot AI delivers largest open-weight AI model ever, as China works around U.S. compute limits →
Tom's Hardware
open-weight-modelschina-ai