Moonshot AI’s July 2026 release of the 2.8-trillion-parameter Kimi K3 open-weights large language model, featuring a one-million-token context window and native multimodal support, established the current peak Kimi score on Humanity’s Last Exam at 46.9–56% across independent leaderboards, with tool-augmented variants reaching the higher end. This places Kimi K3 competitively among open models but behind leading closed systems such as Claude Opus 5.5 and Fable 5.1 at 61–65%+. No subsequent Kimi releases have occurred through late October 2026, and China’s proposed export controls on advanced models add uncertainty to any follow-on development. Traders see limited room for substantial gains without major architectural advances, focusing instead on possible year-end fine-tuning, agentic tool-use optimizations, or community adaptations of the open weights as the main remaining catalysts before the December 31 resolution.