Moonshot's Kimi K3.1 Pro approaches Claude Opus 5.5 on the Artificial Analysis Intelligence Index

Moonshot released Kimi K3.1 Pro, which scores 55 on the Artificial Analysis Intelligence Index, rising to fourth place, only 3 points below the top-ranked Claude Opus 5.5 (58 points). The model improves by 11 points over Kimi K3, with the largest gain on agentic tasks, reaching 1,846 Elo on GDPval-AA. Moonshot says the hallucination rate on Kimi Hallu Bench Hard dropped from 23.4% to 8.3%, and it plans to open the weights on November 4, 2026; the open version's training has had cyberattack data removed, and its solve rate on xdev-bench-hard dropped from 90.5% to 81.0%.