New Anthropic models dominate AI rankings as Chinese models strengthen competitive positions

The Arena platform released its weekly ranking of major AI models for week 33 of 2026 (August 10-16). Anthropic introduced multiple models from the claude-opus-4 series that occupy top positions in the overall standings, with claude-fable-5 holding first place and two newly released models completing the podium with scores within measurement error margins. Chinese models continued their upward trajectory, with kimi-k3-max climbing two positions to 12th overall and maintaining second place in frontend development, while qwen3.8-max made its debut in the intelligent agents ranking within the top 10. Google also achieved notable results with its new gemini-3.7-flash-high model reaching 9th position. In specialized code and frontend development rankings, Anthropic models maintained dominance, while Chinese models secured competitive positions across key categories. In the agents ranking, kimi-k3-max holds fifth place globally, and qwen3.8-max debuted with impressive performance in its first week. Overall analysis indicates Chinese model developers continue narrowing the performance gap with leading international models, particularly in specialized tasks.