Arena AI Model Ranking: Anthropic Leads, Chinese Models Show Gains
Anthropic's new models dominate the latest Arena AI model ranking, securing top spots in the overall leaderboard. Chinese model Kimi-K3-Max maintains a strong position, especially in frontend development.
The latest AI large model blind test ranking from the Arena platform (arena.ai) for the week of August 10-16 shows Anthropic's models heavily leading the overall chart. Several new models from Anthropic's Claude-opus-4 series occupied the top five positions.
The Chinese model Kimi-K3-Max, developed by Moonshot, climbed two spots to 12th place on the overall leaderboard. It also maintained its second position on the frontend development ranking. Alibaba Cloud's Qwen3.8-Max entered the top 10 for intelligent agents for the first time.
Overall, Chinese models are positioned in the upper-middle tier of the rankings. In specialized areas like frontend development, they are demonstrating a strong push to narrow the gap with leading international models. The ranking is based on user votes on the Arena platform, reflecting subjective user preferences.
On the overall list, Anthropic's Claude-fable-5 secured the top spot with 1506 points, followed closely by new Claude-opus-4 series models. Google's Gemini-3.7-flash-high made a notable debut, entering at 9th place.
In the coding model rankings, Anthropic also dominated the top positions. Kimi-K3-Max held sixth place, remaining the highest-ranked Chinese model in this category. Alibaba Cloud's models were also present in the upper-middle section of the coding list.