Tencent has launched Hy4, a new artificial intelligence model with a strong emphasis on coding, game development, financial analysis, and scientific research. Whether you use it as a research agent or for a simpler task, each model output needs review.
But you don’t need a specialized workflow to find artificial intelligence useful.
For everyday tasks like summarizing an article, drafting an email, or explaining an unfamiliar topic, you can also use Wave Browser’s AI Assistant directly from the desktop Side Bar on Windows and Mac.
Here’s what makes Hy4 worth knowing about and where its capabilities stand.
What Is Tencent Hy4?
Released on November 4, 2026, Hy4 is Tencent’s latest large language model (LLM)—a type of artificial intelligence designed to understand requests, work with information, and complete tasks that involve several steps.
Compared with Hy4 Preview, the new model highlights these improvements:
- Understanding images: Hy4 can interpret pictures, analyze charts, and work through math problems presented visually.
- Following instructions: It is designed to follow detailed requests more closely.
- Working with longer documents: It can use information from lengthy materials to help answer questions and complete tasks.
- Handling longer projects: It offers improved support for tasks that require planning, using tools, and checking results.
- Using software: It can interact with software through on-screen controls and typed commands.
These improvements could make Hy4 useful for more demanding work, but Tencent acknowledges that it still trails leading models on some advanced coding and research tests. The best way to judge its usefulness is to try a task you know well and check the results against reliable sources.
What is the Hy4 Preview?
Hy4 Preview was an early version of Tencent’s Hy4 artificial intelligence model, released in August 2026. It let users try its coding, office work, and scientific research capabilities before the full Hy4 release in November 2026.
Understanding Hy4’s Token Pricing
Hy4 is available through Tencent Cloud TokenHub, where pricing depends on how much text it processes and generates.
- Input tokens are the small pieces of text in your questions, instructions, or documents.
- Output tokens are the text it produces in response.
Tencent lists prices per million tokens, with a lower rate for cached input tokens, which are previously processed content that can be reused. Longer documents and more detailed answers generally use more tokens, increasing the cost.
Alongside pricing, the Tencent Hy team reported a pass rate of 96.15% in workplace evaluations. This reflects performance on the tasks tested, rather than a guarantee that every response will be correct.
How to Use Tencent Hy4?
You can try Hy4 through Tencent tools such as WorkBuddy and CodeBuddy.
- WorkBuddy provides access for tasks like preparing documents, organizing project information, and creating presentations.
- CodeBuddy focuses on coding, helping developers write, review, and fix code.
Hy4 also introduces the ability to operate software by clicking buttons, navigating menus, and entering commands. What you can actually do depends on the application you use and the access it provides.
For everyday browsing needs, an AI browser assistant or AI browser extension offers a simpler way to ask questions or summarize information alongside a webpage. Hy4 is the technology behind supported Tencent applications, rather than a browser assistant or extension you install.
Software Engineering and Game Development
The Hy team worked with Tencent’s game developers to improve Hy4’s ability to write code and build better game engines and features. It supports development with tools such as Unreal Engine, Unity, and Godot, helping with complex tasks like creating game rules and multiplayer features.
Hy4 also received training involving professional software such as Blender for creating three-dimensional designs. Users can access the model through Tencent products, including CodeBuddy for coding support, or through Tencent Cloud to connect it to their own applications.
Explore our guide to AI coding tools for more ways artificial intelligence can help with writing, reviewing game logic, and fixing code.
Financial Analysis
Hy4 is designed to help professionals review financial documents, compare information, prepare research, and carry out data analysis. Tencent says it earned the highest score among the open models tested in an assessment involving 213 financial professionals and 579 financial tasks.
The average performance score suggests potential for supporting complex research questions, but evaluation scores do not guarantee accurate answers. Results also depend on the quality and relevance of its training data, so users should still check calculations, sources, and publication dates.
Scientific Research and Safety
Hy4 can support scientific research by helping researchers write code, analyze information, and work through problems in mathematics and other fields. Researchers still need to check its methods and conclusions.
Tencent describes responsible development as a core commitment as it explores the boundaries of AI performance. With Hy4 supporting a broader range of tasks, including coding and business logic, the company says it focused cybersecurity training on finding and fixing software flaws. It also assessed biological safety risks and reported limitations in the model’s ability to plan experiments.
Tencent also announced plans to release Hy4’s model weights and system card, providing further documentation on the model and its safety testing. These assessments describe the company’s findings, rather than a guarantee of safe or accurate results in every situation.
How Tencent Tested Hy4’s Research and Security Skills
Tencent used evaluation benchmarks and expert reviews to assess Hy4’s foundational capabilities. Security testing included binary deobfuscation, which means making deliberately obscured software code easier to understand. The company says training focused on defensive security data to support finding and fixing software flaws. It also reported limitations in experimental design, or planning scientific experiments.
Separate workplace tests placed Hy4 in first place in nine of twelve mainland Chinese office scenarios. However, strong results do not prove a deep understanding of expert workflows in every situation.
Treat these assessments as the first layer of review. When using Hy4 for research, compare findings from diverse sources, check the original information sources, and verify publication dates so you can identify out-of-date information.
Bring AI Into Your Browsing
Hy4’s release highlights how artificial intelligence can support both specialized projects and everyday work. Whether you’re exploring these tools for the first time or finding ways to improve your own work, start with a clear task: understanding a topic, preparing a draft, or organizing ideas for a project.
For everyday browsing, Wave Browser’s AI Assistant lets you ask questions, summarize information, and get writing support directly from the desktop sidebar. It can also help you prepare for collaborative tasks, such as drafting discussion points for a team meeting or turning project notes into a clearer outline to share with colleagues.
Our AI browser guide explores how different browsers bring these tools into your routine. For a closer look at how Wave combines AI assistance with other browsing features, read TechWalls’ Wave Browser overview.
