Articles

The article details Kunlun Wanwei's latest open-source Skywork UniPic 2.0, an integrated multimodal model. This model addresses the high deployment barriers and slow inference speeds of existing large multimodal models by employing a lightweight 2B parameter image generation module to achieve unified image understanding, text-to-image, and image editing capabilities. Through numerous empirical test cases, the article demonstrates UniPic 2.0's robust performance in geographic recognition, abstract understanding, object detection, complex text-to-image generation, and various image editing tasks, highlighting its superior performance compared to similar models with larger parameter counts. The technical reveal section briefly explains its innovative design based on the SD3.5-Medium architecture, integrating multimodal models, and incorporating reinforcement learning. As a highlight of Kunlun Wanwei Technology Release Week, the comprehensive open-sourcing of UniPic 2.0 aims to promote the widespread adoption and application of multimodal AI.

The article provides a detailed review of Agnes AI, an "All in One" AI-driven collaborative workspace. Through personal experience, the author demonstrates its core functions, including Deep Research, which can efficiently generate well-structured, detailed, and traceable research reports; AI Design, which can create visual illustrations that match the content; and AI Slides, which can quickly generate illustrated HTML-based PPTs. The article particularly emphasizes Agnes AI's team collaboration capabilities, supporting real-time/asynchronous editing by multiple users. In addition, it introduces the upcoming Wide Research function, which can schedule hundreds of agents to process complex, large-scale data tasks in parallel and has demonstrated extremely high efficiency and accuracy in actual tests. The article concludes by introducing Agnes AI's developing company, Sapiens AI, and its team background, and elaborates on its vision of an AI-driven workspace, which means transforming AI from content generation to collaborative creation.