Kimi K2.6 Wins AI Coding Contest Against Major Language Models

May 3, 2026

AI Summary

Kimi K2.6, developed by Moonshot AI, emerged victorious in an AI Coding Contest, outperforming notable models like GPT-5.5 and Claude Opus 4.7. The contest involved a Word Gem Puzzle where models scored points based on the length of valid English words formed on a grid.

The AI Coding Contest featured ten language models competing in a real-time programming task called the Word Gem Puzzle.
Kimi K2.6 won with 22 match points, followed by MiMo V2-Pro in second place and GPT-5.5 in third.
The puzzle involved sliding letter tiles on a grid to form valid English words, with scoring based on word length.
Kimi K2.6 utilized an aggressive sliding strategy, while MiMo V2-Pro and Claude did not slide, impacting their performance on larger grids.
The contest revealed significant differences in model performance, particularly on larger 30×30 grids where Kimi's strategy proved more effective.
The results indicate a narrowing gap in capabilities between open-weight models and those from established labs, with Kimi K2.6 scoring 54 on the Artificial Analysis Intelligence Index, close to GPT-5.5's score of 60.
The contest highlighted the importance of real-time decision-making and adaptability in AI models, as static strategies struggled on more complex tasks.

codinggpt-5.5geminiclaudeperformance

Kimi K2.6 Wins AI Coding Contest Against Major Language Models

Related Stories

Thinking Machines Lab develops AI model for simultaneous conversation

ChatGPT Sees Increased Adoption Among Older Users in Early 2026

Optimizing Matrix Multiplication for Swift in LLM Training

arXivLabs Encourages Collaboration on New Features with a Focus on Privacy