Tencent Releases Hy4, Beating GPT in a Computer-Use Benchmark

Tencent has introduced Hy4, a new model in its Hunyuan family that adds visual understanding and the ability to operate computers through both command-line and graphical interfaces.The model is designed for software development, workflow automation, document processing, and interface creation. It can analyze images and charts, interact with applications, and complete multi-step tasks. Its context window supports up to 1 million tokens.

Tencent has introduced Hy4, a new model in its Hunyuan family that adds visual understanding and the ability to operate computers through both command-line and graphical interfaces.The model is designed for software development, workflow automation, document processing, and interface creation. It can analyze images and charts, interact with applications, and complete multi-step tasks. Its context window supports up to 1 million tokens.

Benchmark Results

According to Tencent's evaluations, Hy4 achieved the following results compared with GPT-6.1 Sol and Claude Opus 5.5:

On OSWorld 2.1, a benchmark that evaluates how well AI agents perform tasks in real operating systems, Hy4 outscored GPT-6.1 Sol by 3.7 points and came within 2.9 points of Claude Opus 5.5.However, the model still trails both competitors on ProgramBench, which evaluates programming capabilities, and Terminal-Bench 4.0, which measures performance on command-line tasks.These results are based on Tencent's own evaluations and have not yet been independently reproduced. The OSWorld 2.1 results use partial task scoring, meaning Hy4's score of 78.9 does not represent a 78.9% full task completion rate.You can also try Tencent Hy4 on BotHub for coding, document analysis, and other AI-powered tasks.