Tencent drops Hunyuan Hy4 with computer-use and visual reasoning
Tencent just released Hunyuan Hy4, upgraded with vision and computer-use capabilities to run complex workflows across apps like Blender and CAD. It tackles knowledge work like financial modeling and 3D scientific reconstruction, with open weights coming in days.
AI writing code is old news, but an AI opening up complex 3D software to build scenes by itself is a different story.
Tencent has officially released Hunyuan Hy4, turning the model into a multimodal agent packed with visual reasoning and direct computer-use capabilities.
Hands-on with Blender, CAD, and science
Hy4 can interpret satellite imagery and data charts while driving apps through a mix of graphical user interfaces and command-line inputs. Tencent specifically trained it on specialized tools like Blender and CAD.
In demonstrations, the model handled Blender to reconstruct an entire building, used Three.js to render detailed 3D environments, and ran financial modeling and equity research. In one standout scientific case, it took just 90 minutes to reconstruct a 3D mouse embryo from 7 million cells.
Developers can already access the model via API or through Tencent's free WorkBuddy agent framework, with model weights slated to open-source in the coming days.
Why it matters
If Tencent follows through with open weights, it gives teams a serious local alternative for computer-using agents, without relying entirely on closed Western APIs.
Targeting tough software like Blender and CAD shows a smart focus on real professional workflows over generic chatbot tasks. The real test will be error tolerance: computer use gets brittle fast over multi-step jobs, so we will see how reliable it actually is once the community gets the weights.
Sources (1)
- hy.tencent.ai(opens in a new tab)Originalhy.tencent.ai