Tencent Launches Hy4 AI Model With Computer-Use Skills
Tencent launched Hy4 on November 4, a 770-billion-parameter AI model that can read charts and documents and operate software, with limits on cyber misuse.
Tencent launched Hy4 on November 4, its new flagship AI model with vision and computer-use skills. The 770-billion-parameter model can read charts and documents, reason across several images at once, and operate both command-line tools and graphical software. It is available across Tencent products and through third-party platforms including OpenRouter.
What Hy4 can do
Hy4 adds visual understanding to Tencent's flagship model line. It can interpret charts and documents, compare information across multiple images, and carry out tasks that span several programs. Tencent gave it targeted training for applications such as Blender and CAD software, so it can work inside specialist tools rather than only answering in chat. The model can also operate a computer in a practical sense, using command-line tools and graphical interfaces to complete multi-step work. Tencent says development focused on office tasks, financial analysis, and game development, with product teams helping to train and test the model against real workflows.
Built for office work first
Tencent has integrated Hy4 into WorkBuddy and CodeBuddy, its workplace and coding assistants. Users get a limited free usage quota from November 4 to November 15, and a lower-cost Hy4 flash model is also planned. The company linked the launch to its wider office push: Hy4 can connect chat records, email, documents, schedules, and meeting information through the WeCom suite. Standard API pricing stays the same as the Hy4 preview, at $0.834 per million input tokens and $2.501 per million output tokens, with a 50% discount on Tencent Cloud TokenHub through December 31.
Safety limits come first
The launch stands out for what Tencent deliberately left out. The company restricted the model's cybersecurity training to defensive uses, such as finding software flaws and helping fix them. It excluded material on building hacking tools, breaking into systems, and testing systems for weaknesses. Tencent said its internal security specialists tested Hy4 and found no sign that it posed higher risks in helping users break into software or bypass security protections. Its testing also found the model was not capable of meaningfully making harmful biological experiments easier to carry out. The company plans to release the model weights later, along with a report describing its safety testing. The timing matters. Concern about AI-powered hacking is rising fast, and Tencent disclosed last month that AI agents it used to test third-party models had shown unintended web-browsing behaviour. Tencent said that activity did not come from its Hunyuan models, and it promised stronger controls over agents' internet access and monitoring. Beijing-based Moonshot AI took a similar step last month, limiting its newest model's cyber-offensive capabilities.
Why it matters
Hy4 shows where the AI race is heading: models that do not just answer questions, but see documents, use software, and work across a whole office day. It also shows a shift among Chinese AI firms toward publishing safety limits in public, not just capability claims. For users, the practical test will be whether Hy4's computer-use skills hold up on real office tasks, and whether its open weights arrive with the promised safety report.
Sources
- TechNode - https://technode.com/2026/11/04/tencent-launches-hy4-with-vision-and-computer-use-capabilities/ - TimesLIVE (Reuters) - https://www.timeslive.co.za/news/sci-tech/2026-11-04-tencent-releases-new-ai-model-pledges-safeguards-against-cyberattack-use/