Tencent released the final Hunyuan Hy4 model with image understanding and the ability to operate graphical interfaces and command-line tools within a task. It follows an August preview that lacked those capabilities. The developer describes training for design and game-development workflows, while the final release’s downloadable weights and system card remain forthcoming.
Chinese technology company Tencent has released Hunyuan Hy4 with vision and computer-use capabilities. The final model follows the preview introduced in August.[1], [2]
Within an assigned task, the model can move between graphical interfaces and command-line tools. Tencent describes uses including reading detailed images, interpreting complex charts and working through mathematical questions containing images.[1]
Tencent says it trained Hy4 for workflows involving Blender, CAD design tools, Unreal Engine, Unity and Godot. These cover design and game-development work. The company acknowledges continuing gaps against leading closed models in difficult mathematics, extended development tasks, model iteration and visual question answering.[1]
The August preview had downloadable weights under the Apache 2.0 licence on Hugging Face, a platform for sharing machine-learning models. For the final release, Tencent says it will publish the weights and a system card later. Its assistant and workplace applications are also part of the distribution plan, including the WorkBuddy and CodeBuddy tools for office and software-development tasks.[1]
