9B announced the launch of the "AI Infinite Arena" large model evaluation ranking list, and simultaneously released the first round of model rankings. Among the 10 UP masters' evaluations, GPT-6 Astra took the top spot, defeating GLM-5.3 to win the "most number of times champion"; in the top five, three are domestic large models.
According to B station, "AI Infinite Arena" is a competitive platform that collects UP master AI large model evaluations. It consists of real scenario tests of hundreds of large models by UP masters from various fields, covering evaluation topics such as code, reasoning, collaboration, and knowledge.
No Topic Restrictions, Let Real Workflows Speak
Different from traditional benchmarking tests, B station's "AI Infinite Arena" has no restrictions on topics or evaluation dimensions - UP masters from various sections use real workflows, professional applications, and fun ideas as self-proposed topics, intuitively demonstrating the actual performance differences of each model in the same topic competition.
The ranking list includes comparative evaluations of mainstream models such as DeepSeek, Kimi, ChatGPT, Claude, Gemini, Douba, Qwen, Hy, and MiniMax. The rankings are updated in real time and continue to open registration for all UP masters on the platform.
Public information shows that over the past year, the consumption time of AI knowledge content on B station increased by 72%, with more than 190 million monthly users watching AI content. A large amount of evaluation content covering different fields, dimensions, and topics has emerged on the platform, forming a content ecosystem from release to discussion, evaluation, use, help, and co-construction.
