Tencent launches Hy4 with vision and computer use
The low-cost model works across GUI and CLI tasks, while open weights and the full system card are still promised rather than available.
Tencent has released Hy4, its first model in the family with visual perception and computer-use capabilities. The company says it can work through both graphical interfaces and command-line tools, including specialized software such as Blender and CAD packages.
Tencent is positioning the model for frontend development, games, finance, office work and research. It reports gains over the Hy4 preview in internal game-development evaluations and says a finance review involving 213 professionals placed Hy4 ahead of several open-weight rivals. Those results have not yet been independently reproduced.
The model is available through Tencent Cloud, OpenRouter and Tencent products. API pricing starts at 0.042 dollars per million cached input tokens, 0.834 dollars for regular input and 2.501 dollars for output.
Despite Tencent calling Hy4 an open model, the weights are not available yet. The company says it plans to release them together with vLLM and SGLang support and a full system card. It also acknowledges that a notable gap remains versus top proprietary systems.
The launch matters because it brings another low-cost Chinese model into multimodal agent workflows, where DeepSeek, Kimi and GLM are already competing.
Tencent
