Tencent releases official version of Hy4… supports visual understanding and computer manipulation

Tencent released the full version of its Hy4 large language model on June 4, adding visual understanding and computer-control capabilities for the first time. The model can parse images and complex charts, handle visual reasoning in diagram-based math problems, and operate software through both command-line and graphical interfaces, including professional tools such as Blender and CAD. Tencent said the release also improves instruction-following, long-context comprehension, and multi-step agentic task performance. The launch matters because it pushes Tencent's in-house model deeper into multimodal and agentic use cases, where it will compete with rival Chinese and global AI systems. Next, watch for pricing and API availability, enterprise adoption, and independent benchmarks of its visual and computer-use performance.