Matthew Berman tests Kimi K3.1 on the Rubik's Cube benchmark, showing it pre-calculate and play out a full solve on a scrambled cube.
Continue to AI University →