荣耀发布MagicGUI自研开源大模型，首发搭载折叠屏Magic V5

Core Insights - Honor officially launched the MagicGUI self-developed multimodal perception model during the 2025 World Artificial Intelligence Conference (WAIC) [1] - The MagicGUI model features a parameter scale of 7 billion, supporting multimodal perception and automatic execution planning capabilities [1] - The model has achieved industry-leading performance in various tasks, including visual positioning, page understanding Q&A, single-step operation, and action sequence execution [1] Company Highlights - The MagicGUI model is the first open-source GUI large model introduced by Honor [1] - The model's accuracy in common scenarios with the Honor Magic V5 reached 91.5%, which is a 16.4% improvement over top industry open-source models [1] - The launch of this model is expected to accelerate the application of AI in the smartphone sector [1] Industry Context - The evaluation results indicate that the MagicGUI model outperforms competitors in mobile visual positioning and other key tasks [1] - The advancements in AI capabilities showcased by Honor may set new benchmarks for the industry [1]