The new issue of the Arena AI Large Model Weekly Rankings has been officially announced, with major manufacturers engaging in a new round of competitive showdowns across multiple sub-ranks such as comprehensive, code, agents, and multimedia. In this ranking, the newly launched model from Anthropic showed remarkable performance, while domestic large models maintained stable tiered performances in various specialized fields.

In the highly anticipated agent ranking, Claude Opus5.5 (Max) from Anthropic demonstrated an overwhelming advantage upon entering the ranking, securing the top position immediately. Its task confirmation success rate reached 17.29%, and its tool hallucination rate was controlled at an ultra-low level of 0.37%. Meanwhile, domestic large models steadily advanced in the agent category, with multiple products such as DeepSeek V4.1Flash (Max), Kimi K3 (Max), Tencent Hy4preview, and Xiaomi MiMo V2.6Pro all ranking among the top, maintaining a stable mid-to-top tier position.

The top rankings in the comprehensive and code categories also saw fierce competition. In the comprehensive ranking, Google's gemini-4-argon-high continued to hold the top position, while multiple main models from Anthropic occupied several of the top four positions. In terms of domestic models, Moonshot's kimi-k3-max remained in 16th place, while Alibaba's qwen3.8-max, Xiaomi's mimo-v2.6-pro, and Zhipu's glm-5.3-max also maintained stable rankings on the list. In the code ranking, gemini-4-argon-high continued its strong performance, leading the list, while Claude Opus5.5 (High) significantly improved its ranking and entered the top five. The code processing capabilities of Moonshot and Xiaomi's domestic models also remained in the industry's forefront.

In niche areas such as front-end development, multi-modal image generation, and video production, the technical iterations of various manufacturers have not slowed down. In the front-end development ranking, Claude Opus5.5 Max continued to lead due to its excellent interaction and rendering capabilities. Xiaomi's newly added mimo-v2.6-flash and pro versions both showed strong technical potential. In the AI video and image generation rankings, although Google and OpenAI occupied some top positions, multiple key generation models from domestic teams such as ByteDance, Alibaba, and MiniMax also performed stably, demonstrating strong global competitiveness. As large models are deeply applied in various scenarios, the market is expected to see more version updates and product upgrades next week.