The new weekly ranking of Arena AI large models was officially released, and various manufacturers engaged in a new round of strength competitions across multiple categories such as comprehensive performance, code, AI agents, and multimedia. In this ranking, the new model launched by Anthropic performed extremely well, while domestic large models maintained a solid performance in various niche areas.
In the highly anticipated AI agent rankings, Claude Opus5.5 (Max) from Anthropic emerged with an overwhelming advantage right from its debut on the list, taking the top spot outright. Its task confirmation success rate reached 17.29%, and the tool hallucination rate was kept at an extremely low level of 0.37%. Meanwhile, domestic large models have been steadily making progress in the AI agent field. Products such as Deepseek V4.1Flash (Max), Kimi K3(Max), Tencent Hy4preview, and Xiaomi MiMo V2.6Pro have all ranked among the top, maintaining a stable mid-to-upper tier position.
The top positions in both the general and code rankings are highly competitive. In the general ranking, Google’s gemini-4-argon-high continues to hold the top spot, while several main models from Anthropic follow closely behind, taking multiple positions in the top four. Among domestic models, kimi-k3-max ranks 16th, while阿里 qwen3.8-max, Xiaomi mimo-v2.6-pro, and Zhipu glm-5.3-max also maintain stable rankings on the list. And in the code ranking, gemini-4-argon-high continues to lead strongly, while Claude Opus5.5 (High) has surged significantly into the top five. The code processing capabilities of Moon’s Dark Side and Xiaomi, among other domestic models, also remain at the forefront of the industry.
In front-end development, as well as niche areas such as multi-modal image generation and video, the technological advancements of various manufacturers have not slowed down. On the front-end development list, Claude Opus5.5Max continues to lead due to its excellent interaction and rendering capabilities. The newly listed versions of mimo-v2.6-flash and pro by Xiaomi both demonstrate impressive technical strength. In the AI video and generative charts rankings, although Google and OpenAI occupy some top positions, several major generation models from domestic teams such as ByteDance, Alibaba, and MiniMax also perform strongly, showing strong global competitiveness overall. With the deep implementation of large models in various niche scenarios, the market is expected to see more version updates and product enhancements next week.