Kimi K3, an open-weight AI model, has released benchmarks indicating its performance is comparable to frontier models like 5.6 Sol and Fable. This release suggests that frontier AI companies are not pulling away from open-weight models, with the performance gap between public-use models now approximately five weeks.