Tencent releases Hy4, a frontier LLM with computer use and visual understanding

Tencent has introduced Hy4, a large language model that adds computer-use capabilities for the first time, built on visual understanding, reasoning and tool-calling for long-horizon agentic work. The model has 770 billion parameters, 49 billion of them active, and a 1 million token context window. Tencent says it improved on its Hy4 Preview, released August 28, in instruction following, long-context understanding, agentic tasks and deep research.

Tencent says the model is stronger at front-end and game development, reporting gains of 19.1 percent on the GameDev bench and 9.8 percent on GameLogic over the preview. It did not publish the baseline figures behind those numbers. The company also says the model was integrated with Unreal Engine, Unity and Godot.

Pricing is unchanged at about $0.90 per million input tokens, $2.70 per million output tokens and $0.045 per million cached input tokens. Hy4 is available on Tencent Cloud TokenHub, WorkBuddy, CodeBuddy and Yuanbao, and through OpenRouter, Hermes and OpenCode. Tencent plans to release the weights and system cards in the coming days and a cheaper Flash version soon.

Developers can try a computer-use model with a 1 million token context through hosted endpoints such as OpenRouter and OpenCode at the listed per-token prices. Tencent says those prices match the preview, so budgets built around the earlier version should carry over. Once the promised weights are published, teams will be able to evaluate the model independently and consider self-hosting it.

Tencent's benchmark gains are self-reported and the baselines are not public, so teams should run their own evaluations before switching agent workloads to the model.