Tencent Hunyuan Hy4 Preview Officially Launched

Today, we are releasing Hy4 preview, with a total of 770 billion parameters, 49 billion activated parameters, and a context length of 1 million tokens, demonstrating exceptional capabilities in real-world productivity tasks such as coding, office work, and scientific research.

Next-Generation Flagship Model

Hy4 preview has undergone significant expansion in model size, context length, and data scale. Joint advancements in pre-training and post-training have driven another major leap in intelligence levels, firmly establishing it in the top tier of open-source models.

Built for Productivity

By co-developing high-quality datasets with top experts from Tencent in fields such as software engineering, gaming, finance, and security, Hy4 preview has achieved significant progress across various real-world productivity tasks:

Software Engineering:Enhanced capabilities in understanding, planning, debugging, and verification for long-horizon development tasks, further improving visual aesthetics and interaction quality in front-end development;

Office Analytics:Significantly improved understanding of complex office environments and financial analysis capabilities, with a focus on optimizing data analysis and cross-file collaboration to complete the full workflow from information processing to the delivery of documents, spreadsheets, and presentations;

Game Development:Enhanced ability to directly generate playable prototypes from single-sentence requirements, along with proficient use of game engines, enabling developers to continuously refine complex game projects through multi-turn interactions;

Scientific Research:Significantly improved capabilities in understanding, reasoning, and solving complex scientific research problems, with substantial progress made by the model in various scenarios including AI R&D, molecular dynamics simulations, condensed matter physics, and fundamental mathematics.

Meanwhile, Hy4 preview continues to integrate deeply with products such as CodeBuddy and WorkBuddy to optimize the real-user experience in productivity scenarios.We organized 163 internal experts to conduct blind tests on 203 engineering tasks. The results show that Hy4 Preview (average score: 2.99/4.00) slightly outperformed GLM-5.3 (average score: 2.92/4.00; win rate: 46.8% / draw rate: 12.8% / loss rate: 40.4%) and Kimi K3 (average score: 2.94/4.00; win rate: 51.2% / draw rate: 7.9% / loss rate: 40.9%).

Hy4 Preview is now open-source, available on Tencent Cloud TokenHub and OpenRouter, and can be experienced through products such as WorkBuddy/CodeBuddy, Yuanbao, and Ima. We remain committed to openness and high cost-effectiveness, making model advancements accessible to a broader user base. To further collect user feedback and enhance model performance, WorkBuddy/CodeBuddy will launch a two-week limited-time free access campaign for users.

Since rebuilding its infrastructure in February, the Hunyuan large language model has released major version updates every two months on average. By adopting a strategy of releasing preview versions first followed by official releases, Hunyuan continuously integrates real-world feedback into its R&D process, enabling the model to improve through solving practical problems.

Hy4 Preview is an early iteration of Hy4. Both pre-training and post-training stages still have significant room for improvement, and there are some known issues, such as extended reasoning times for complex tasks and a tendency toward excessive self-verification. We will continue to iterate agilely.

Similar to Hy3 Preview, we aim to release Hy4 Preview as soon as possible to gather extensive real-world feedback, thereby significantly enhancing the official Hy4 release. Meanwhile, we will continue to leverage our unique advantages from deep collaboration with Tencent’s products and experts, steadily improving the accessibility and upper limits of productivity.

192K Views