Tencent Releases Hunyuan Hy4 with Vision and Computer Use Support
Tencent officially released the latest generation of its large language model, Hy4, claiming it ranks among the strongest open-source models. This version significantly expands training scale, introduces vision capabilities for the first time, and supports computer use (GUI/CLI interaction), with major improvements in instruction following, long-context understanding, and complex task execution. Hy4 is now available on Tencent Cloud and OpenRouter, with plans to open-source weights soon and support the vLLM/SGLang ecosystem.
Full Report
IT Home reported on November 4 that Tencent Hunyuan released the official version of its latest large language model, Hy4, today (November 4). The company stated that with a comprehensive leap in model capabilities, Hy4 has entered the ranks of the strongest open-source models for the first time. Model weights will be open-sourced soon, along with the release of a model system card. Tencent is actively promoting more responsible open-sourcing, with a focus on risks such as cybersecurity and biosecurity.
Compared to the Hy4 preview, Hy4 significantly expanded both pre-training and post-training scales, introduced visual capabilities for the first time, and greatly improved overall performance in instruction following, long-context understanding, long-horizon tasks, and deep research.
Hy4 introduces visual capabilities for the first time, enabling it to combine perception, reasoning, and tool calling for tasks such as fine-grained image processing, complex chart analysis, and mathematical visual reasoning. Hy4 also supports computer operation for the first time, allowing flexible integration of command-line interfaces and graphical user interfaces. Tencent has conducted specialized training for professional software like Blender and CAD, as well as cross-application long-horizon tasks, continuously optimizing planning and execution capabilities for professional workflows.
According to the introduction, this model is designed for real-world productivity across various scenarios, including visual and computer tasks, front-end and game development, finance and office environments, scientific research and model R&D, and long-term personal assistants.
As previously reported by IT Home, Tencent released the Hy4 preview on August 28, featuring 770B total parameters, 49B active parameters, and a context length of 1M.
At the time, officials stated that the Hy4 preview had significantly expanded in terms of model size, context length, and data scale, with improvements in both pre-training and post-training leading to another major leap in intelligence levels.
Currently, Hy4 is available on Tencent Cloud and OpenRouter, and can be experienced through products such as CodeBuddy / WorkBuddy, Yuanbao, and ima. It delivers significantly better performance than the Hy4 preview while maintaining the same API pricing.
The company plans to open-source the model weights and has already collaborated with the vLLM/SGLang communities and Inferact Inc. to provide support for the open-source ecosystem. A model system card, including more comprehensive safety evaluations and other details, is also scheduled for release at that time.
Although Hy4 is currently one of the strongest open-source models, there remains a significant gap compared to top-tier closed-source models, particularly on high-difficulty test sets in areas such as advanced mathematics, long-horizon development, model iteration, and visual question answering built internally. Tencent Hunyuan stated that they will continue to steadily expand pre-training and post-training scales, striving to narrow the gap with leading closed-source models.
Hy4 flash is also planned for release soon. It will significantly improve cost-performance ratio and inference speed, enable more frequent collaborative iterations with products, and continuously enhance practicality.