Tencent Hy4 Launch: 770B Vision Model With Computer Use

Tencent officially launched the full Hy4 model on November 4, 2026, delivering a 770-billion-parameter mixture-of-experts architecture with visual understanding and native computer-operation capabilities. The release marks a significant expansion beyond the Hy4 preview, which shipped without vision or computer-use features, and positions Tencent's open-weights flagship as a multimodal competitor to models like Claude and GPT-4.

Model Architecture and Scale

Hy4 operates as a 770B-parameter mixture-of-experts system, activating only a subset of parameters for each inference task to balance performance with efficiency. The model supports a 1-million-token context window, enabling it to process book-length documents, multi-hour meeting transcripts, and extensive codebases in a single prompt. Tencent expanded both pre-training and post-training data volumes compared to the preview version, resulting in improved instruction-following accuracy and long-range reasoning across tasks that require multi-step planning or extended context recall.

Visual Understanding Capabilities

For the first time in the Hunyuan line, Hy4 ships with native visual understanding. The model can analyze images, diagrams, screenshots, and charts, then answer questions or generate text based on visual content. This multimodal capability enables use cases like technical documentation generation from architectural diagrams, accessibility descriptions for images in content management systems, and visual question-answering workflows in customer support tools. Tencent integrated vision directly into the core model rather than bolting on a separate vision encoder, allowing Hy4 to reason about text and images in a unified latent space.

Computer-Operation Skills

Hy4 introduces native computer-use agent functionality, letting the model interact with graphical user interfaces, execute commands, and control software applications. The system can navigate web browsers, operate desktop applications, and automate multi-step workflows that require clicking buttons, filling forms, or reading on-screen text. Early demonstrations show Hy4 completing tasks like booking appointments through web interfaces, extracting data from spreadsheets, and managing email workflows inside Tencent's WorkBuddy and WeChat Work platforms. This positions Hy4 as a direct competitor to Anthropic's Claude computer-use capabilities and Google's emerging agent frameworks.

Deployment and Pricing

Officially launched on November 4, 2026, Hy4 is available across Tencent's product suite, including Tencent Cloud TokenHub, WorkBuddy, CodeBuddy, Yuanbao, and the WeChat Work intelligent assistant Dayuan. Third-party API providers like OpenRouter, Hermes, and OpenCode also integrated Hy4 on launch day. Tencent is offering a 50% discount on API pricing through December 31, 2026, making Hy4 inference cost-competitive with smaller open-source models during the promotional window. The open-weights release allows developers to self-host Hy4 for on-premises deployments, addressing data-residency and compliance requirements in regulated industries.

What This Means

Tencent's Hy4 launch signals China's accelerating progress in frontier open-weights models, particularly in multimodal and agentic AI. By combining 770B-scale language modeling with vision and computer-use skills in a single open architecture, Tencent delivers a credible alternative to closed Western models for developers building multimodal workflows or automation agents. The integration across Tencent's ecosystem and the aggressive launch pricing suggest the company is prioritizing adoption velocity and ecosystem lock-in over immediate API revenue. For enterprises evaluating open-source LLM options, Hy4's computer-operation capabilities and 1M-token context make it a strong candidate for workflow automation and document-processing pipelines, especially in markets where data sovereignty favors on-premises deployments over cloud API calls.