World's First Open-Source Medical Video LLM Launched by UII: A Game-Changer in AI Innovation

Introduction

On April 24, 2026, United Imaging Intelligence (UII) made a significant leap in the field of artificial intelligence by unveiling the world's first ever open-source medical video large language model (LLM) known as uAI NEXUS MedVLM. This groundbreaking tool is set to redefine how medical videos are processed and analyzed, offering unprecedented spatiotemporal accuracy in clinical settings. The introduction of such a robust model invites an international call to action for developers and researchers in the AI community to further enhance its capabilities.

The Innovative uAI NEXUS MedVLM

UII's uAI NEXUS MedVLM emerges as a groundbreaking solution that leverages a monumental dataset consisting of 531,850 pairs of video instructions across eight clinical environments, including robotic surgeries, laparoscopic procedures, endoscopy, open surgeries, and nursing care. Despite having only 4B/7B parameters, this LLM significantly outperforms leading general-purpose models such as GPT-5.4 and Gemini 3.1 in key medical video tasks. Its remarkable accuracy—89.4% in surgical safety assessments compared to 1.8% and 10.1% of its competitors, respectively—demonstrates its unparalleled capabilities.

Additionally, it achieves a spatiotemporal action localization score that is up to 14 times higher than GPT-5.4 and four times higher than Gemini 3.1. Furthermore, MedVLM excels in video report generation, receiving a score of 4.2 out of 5, again surpassing the performance of its well-known counterparts.

A Global Open Challenge for Collaborative Innovation

To accelerate the development of machine learning models for medical video applications, UII has initiated the phased release of its MedVidBench dataset, launching with 6,245 rigorous benchmark test samples as open source. This initiative, comprising eight diverse surgical datasets, represents a monumental global achievement in both scale and clinical precision. Developers are now invited to evaluate their models through a unified ranking system, which facilitates automatic comparisons against private benchmark data. The results feed into a continuously updated global ranking, enabling transparent performance assessments across different models.

UII urges researchers, developers, and AI healthcare institutions worldwide to join this open challenge and contribute to advancing AI applications in medical video through collaborative innovation.

For more details and to participate, the project can be accessed at MedGRPO Project.

Overcoming Historical Challenges in Medical Video Understanding

Understanding medical videos has long been considered one of the most formidable challenges in AI, requiring microscopic spatial reasoning, complex temporal logic, and unwavering clinical precision. Historically, progress has been hindered by significant limitations in clinical data availability and the high costs associated with expert annotation. However, UII has successfully navigated this obstacle by establishing a vast annotation framework, meticulously mapping critical attributes such as instrument trajectories, spatial positioning, precise surgical actions, and crucial risk indicators in various clinical videos. This unparalleled dataset empowers uAI NEXUS MedVLM with a comprehensive and robust clinical intelligence platform.

Integrating Perception, Reasoning, and Decision-Making
Building on this foundation, the model seamlessly incorporates perception, reasoning, and decision-making capabilities. It provides high-precision spatiotemporal localization of instruments and automated procedural recognition, employing advanced reasoning to convert complex video sequences into structured clinical reports, regional descriptions, and concise workflow summaries. This goes beyond merely passive observation and elevates these insights into active decision-making processes, supporting predictions of subsequent steps, evaluations of surgical proficiency, and comprehensive safety risk assessments.

Transforming AI Innovation into Real Clinical Impact

Designed for clinical implementation, uAI NEXUS MedVLM facilitates more informed decision-making and data-driven quality control across all surgical workflows, while simultaneously reducing the learning curve for healthcare professionals. This innovation enhances both the efficiency and consistency of medical training.

Looking ahead, uAI NEXUS MedVLM has the potential to serve as a central perceptual and cognitive engine for AI systems integrated into the physical healthcare environment. Collectively, these advancements aim to create a closed-loop system of visual perception, cognitive reasoning, and physical execution, paving the way for a more automated, standardized, and intelligent healthcare ecosystem.