ChainCatcher message: Tencent Hunyuan has released and open-sourced a new generation large language model, Hy4 preview. The model has a total of 770B parameters and 49B activated parameters. Its context length exceeds 1M. It demonstrates strong capabilities in real productivity tasks such as coding, office work, and scientific use cases, solidifying its position in the top tier of open-source models. Hy4 preview has been significantly expanded across model size, context length, and data scale. It further improves performance in real-world scenarios through co-building with high-quality data from Tencent experts in software engineering, games, finance, and security, as well as deep collaboration with products like WorkBuddy.

In software engineering, the model strengthens understanding, planning, debugging, and validation capabilities for long-context development tasks. It enables you to build complex frontend projects from scratch, such as a Three.js microcosm town. For game development, it supports generating playable prototypes from a single sentence and completing a full demo in Unity. For intelligent office work, it can handle complex financial audit and review tasks—selecting, analyzing, and delivering outputs from multiple documents. In scientific research, it accelerates tasks such as molecular dynamics simulations and has initially formed a recursive self-improvement loop.

Hy4 preview can be experienced in WorkBuddy/CodeBuddy domestic and international versions, as well as in Tencent products such as Yuanbao and ima. It can also be called via API through Tencent Cloud Tokenhub and OpenRouter. WorkBuddy/CodeBuddy will run a two-week limited-time free promotion. Since the Hunyuan model’s infrastructure reconstruction, it has been updating to a new major version on average every two months. It continuously optimizes using a strategy of releasing previews first, followed by official versions.