An open-weights Chinese model, Kimi K2.6 from Moonshot AI, has outperformed leading Western AI models including OpenAI's GPT-5.5, Anthropic's Claude Opus 4.7, and Google's Gemini in a programming challenge. The AI Coding Contest's "Word Gem Puzzle" saw Kimi K2.6 achieve the highest score by employing an aggressive sliding strategy to reconstruct words on the game board. While Kimi K2.6 and Xiaomi's MiMo V2-Pro secured the top two positions, models from major Western labs like OpenAI and Anthropic placed lower, indicating a potential shift in the competitive AI landscape. AI
Summary written by None from 1 source. How we write summaries →
IMPACT Suggests that specialized models from non-Western labs may be closing the gap or surpassing frontier models in specific task performance.
RANK_REASON This is a report on a third-party coding challenge evaluating existing models, not a release from a frontier lab. [lever_c_demoted from research: ic=1 ai=1.0]