Kimi K2.6 from Moonshot AI Wins Programming Challenge Over Claude, GPT-5.5, and Gemini — 329 Points on HN

ThinkPol / Hacker News·high signal

Moonshot AI's open-weights Kimi K2.6 won a head-to-head programming challenge (Word Gem Puzzle) with a 7-1-0 record and 22 match points, beating GPT-5.5 (third) and Claude Opus 4.7 (fifth). On broader benchmarks, K2.6 scores 58.6 on SWE-Bench Pro, tying GPT-5.5, while Claude Opus 4.7 leads at 64.3. The 329-point HN discussion (185 comments) focused less on the specific challenge and more on the trajectory of Chinese open-weight models closing the frontier gap. K2.6 model weights are publicly available.