3.0. The model tops the LMArena Global Blindness first

According to the official, LMArena's latest vernacular list at the International Large Modeling Stadium shows that mixed images 3.0 out of 26 large models globally, top closed-source models like nano-banana, the result of which comes from "blindness" by all users around the globe. The official presentation stated that LMArena was an innovative AI model assessment platform launched by the University of California at Berkeley, United States, and that the core method of evaluation was based on a “blindness” mechanism based on real human preferences, allowing users to vote anonymously on responses to different AI models and measure their performance. Following user input questions, the Platform provided responses to two models, with users opting for better answers based on preferences, and voting results directly shaping public rankings. Because of the large number of visits to the platform and the closeness of the evaluation mechanism to the actual experience, it is now the most authoritative playing field list at the international level. The hybrid image 3.0 was released on 28 September and is the first industrial-grade multi-modular model with a parameter size of 80B. The current version has already opened up the graphics capability and will be followed up with the graphics, image editing, multi-wheel interaction, etc。