Key points
- Tencent released Hy4 with 770 billion parameters and 49 billion active parameters.
- The model adds computer-use capabilities and deep integration with Unreal, Unity and Godot.
- Performance gains were reported in instruction following, long-context understanding and deep research.
Tencent Holdings Ltd. introduced Hy4, a large language model that adds computer-use capabilities for the first time, according to a report by SiliconANGLE.reported The release marks a shift toward long-horizon agentic performance using visual understanding and tool-calling.Tencent released a new AI model that can use computers and understand visuals.Tencent Hy4 introduces computer-use capabilities to a 770B parameter model with 49B active parameters.
Evidence: reported
Tencent releases Hy4, powerful frontier LLM introducing visual understanding and computer useaifleet-siliconangle
What was released or claimed
The company stated that Hy4 represents a significant performance leap over the Hy4 Preview, which launched on August 28.reported The new model weighs in at 770 billion parameters with 49 billion active parameters.reported Both versions provide a context window of one million tokens.
Evidence: reported
Tencent releases Hy4, powerful frontier LLM introducing visual understanding and computer useaifleet-siliconangle
Evidence: reported
Tencent releases Hy4, powerful frontier LLM introducing visual understanding and computer useaifleet-siliconangle
Who it affects
SiliconANGLE noted that Hy4 targets workplace collaboration, human experience and professional work.reported The company said it optimized the model for web development, document summarization, visual creative tasks and document presentation.
Evidence: reported
Tencent releases Hy4, powerful frontier LLM introducing visual understanding and computer useaifleet-siliconangle
What is still unknown
The material does not specify independent benchmark scores or third-party audits of the claimed performance gains. It is unclear how the model handles safety constraints during complex computer-use tasks.
What changes, what does not
Tencent said it scaled pre-training and post-training substantially. The company identified excessive reasoning and over-verification in the previous version’s thinking processes and stated it addressed these issues.
Methods and limits
The research tests long-horizon agentic performance through advanced visual understanding, reasoning and tool-calling. Reported results include gains in instruction following, long-context understanding and deep research.reported Stated limitations in the prior version included excessive reasoning and over-verification during thinking processes.
Evidence: reported
Tencent releases Hy4, powerful frontier LLM introducing visual understanding and computer useaifleet-siliconangle
Why it matters
- Architects can evaluate Hy4 for front-end and game development tasks involving intricate business logic and state management.
- The one-million-token context window allows comparison with open-weights frontier models from Mistral AI, Moonshot and Z.AI.
- Integration with Unreal Engine, Unity and Godot supports complex game logic and multiplayer implementations.
- Improvements in long-context learning may enhance document summarization and visual creative workflows.
Topics: models, Tencent, LLM, computer-use, game development
Updated
Then / now / next
The Hy4 Preview launched on August 28 with reported issues in reasoning efficiency.
Tencent released Hy4 with computer-use capabilities and enhanced visual understanding.
Monitor independent evaluations of the model’s agentic performance and safety controls.
Story timeline
- Then The Hy4 Preview launched on August 28 with reported issues in reasoning efficiency.
- 04 Nov 2026 Tencent releases Hy4, powerful frontier LLM introducing visual understanding and computer use
- Now Tencent released Hy4 with computer-use capabilities and enhanced visual understanding.
- Next Monitor independent evaluations of the model’s agentic performance and safety controls.
The strongest doubt
Counterpoint from the sources
The material notes that some models struggle with producing stateful, elegant and fully functional user interfaces.
For readers
Developers
- Hy4 supports front-end and game development with complex business logic.
- It integrates with Unreal Engine, Unity and Godot for multiplayer implementations.
Researchers
- The model uses 770 billion parameters with 49 billion active parameters.
- It features a one-million-token context window comparable to other frontier models.
Business
- Tencent claims gains in instruction following and long-context understanding.
- The model targets workplace collaboration and professional work tasks.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
active parameters
Parameters used during inference in a mixture-of-experts model
context window
The maximum number of tokens a model can process at once
agentic performance
The ability to perform long-horizon tasks autonomously
tool-calling
The capability to invoke external tools or APIs
Questions readers ask
What are the parameter counts for Tencent Hy4?
The model has 770 billion total parameters and 49 billion active parameters.
Which game engines does Hy4 support?
It integrates with Unreal Engine, Unity and Godot.
When was the previous version released?
The Hy4 Preview arrived on August 28.
Connected stories
Reading receipt
- Tencent released Hy4 with 770 billion parameters and 49 billion active parameters.
- The model adds computer-use capabilities and deep integration with Unreal, Unity and Godot.
- Performance gains were reported in instruction following, long-context understanding and deep research.
Still open: How does Hy4 perform on independent safety benchmarks for computer-use tasks?
Still open: What are the specific latency metrics for long-horizon agentic workflows?