Search
The AI team at Alibaba Cloud International has unveiled the latest version of their multimodal large model, Ovis2.5, showcasing exceptional comprehension and reasoning skills across various domains including general multimodal benchmarks, intricate chart interpretation, and OCR. This upgraded model excels at addressing functional challenges through image analysis, logistics order and invoice scrutiny, and even scene-based photo location prediction. On the renowned multimodal evaluation platform OpenCompass, Ovis2.5-9B and Ovis2.5-2B secured impressive scores of 78.3 and 73.9, respectively, securing top positions. The model's key innovations include the integration of native resolution visual perception technology and an optional "thinking mode," delivering dual advancements in both performance and efficiency.
-
C114 Communication Network -
Communication Home
