An open-weights Chinese model just beat Claude, GPT-5.5, and Gemini in a programming challenge
Summary
Kimi K2.6, a Moonshot AI open-weight model, won the AI Coding Contest Word Gem Puzzle, beating Claude, GPT-5.5, and Gemini Pro on multiple grid sizes. The article analyzes the scoring rules and strategies, highlighting how open-weight models and local deployment are reshaping AI benchmarking and access.