Chinese AI models surge in weekly rankings as Alibaba's Qwen breaks top 5

Arena's weekly assessment of AI models from July 27 to August 2 puts Anthropic's Claude Fable 5 at the top with 1509 ELO points. The week's major development was Alibaba's debut of Qwen 3.8-Max at fifth position overall, scoring 1496 points and trailing the leader by just 13 points—marking the best performance yet for a Chinese model. Qwen 3.8-Max also entered the code generation top 10 with 1530 points, while Moonshot AI's Kimi K3-Max took eighth place at 1531 points. In image synthesis benchmarks, ByteDance's Seed Dream 5.0 Pro advanced five spots to land sixth with 1257 ELO points. Anthropist's Claude Opus 5 Max continues dominating frontend development at 1705 points, now with half of the top 8 positions held by Chinese competitors. Across the board, Chinese AI models are narrowing the performance gap with international leaders, with substantial gains visible across multiple specialized benchmarks.