Tencent Officially Releases Hy4 Open-Source LLM for Enterprise Productivity
Tencent has officially released the Hy4 open-source large language model, marking its entry into the tier of top open-source models with newly added visual capabilities, computer-use operation features, and expanded pre-training and post-training scales. This release marks a significant milestone for Tencent as it bridges the gap between conversational AI and practical digital workers through computer-use agents and multimodal capabilities designed for real enterprise productivity. By open-sourcing the model weights and offering ecosystem support, it accelerates the adoption of advanced open-source AI in professional fields like software development, finance, and scientific research. Hy4 supports a 1M context length and combines perception, reasoning, and tool invocation for complex visual tasks and GUI/command-line computer operations. It is currently available on Tencent Cloud and OpenRouter, with upcoming plans to release the Hy4 flash version to improve cost-efficiency and inference speed.
## BACKGROUND
Large Language Models (LLMs) rely heavily on pre-training and post-training scales to boost capabilities like instruction following and long-form comprehension. Features like computer-use allow AI models to interact directly with graphical user interfaces and execute cross-application tasks, transforming them from passive chatbots into active digital agents.