Latest Evaluation of Multimodal Large Models Released! Gemini-3-Pro Ranks First with a Major Gap, Doubao and SenseTime Lead the Domestic Group, Qwen3-VL Becomes the First Open-Source Model to Achieve High Scores

According to the latest SuperCLUE-VLM rankings, Google's Gemini-3-Pro scored 83.64 points, showing a significant lead, particularly in visual understanding and reasoning. Domestic models performed outstandingly, with SenseTime's SenseNova V6.5Pro and ByteDance's Doubao ranking second and third, showcasing China's rapid progress in the multimodal field. The evaluation covers three core capability dimensions.