Tencent Releases and Open-Sources Tencent Hy4 Preview
Tencent Releases and Open-Sources Tencent Hy4 Preview
腾讯发布并开源腾讯混元 Hy4 预览版
Ranked among the top tier of open-source models, Hy4 preview is built for real-world productivity tasks, delivering outstanding performance across coding, office work, and scientific research. Tencent has released and open-sourced Tencent Hy4 preview, a next-generation large language model with 770B total parameters and 49B active parameters, and a context window exceeding 1M tokens. It demonstrates outstanding capabilities on real-world productivity tasks spanning coding, office work, and scientific research.
作为开源模型中的佼佼者，Hy4 预览版专为实际生产力任务而构建，在编程、办公和科学研究领域表现卓越。腾讯正式发布并开源了腾讯混元 Hy4 预览版，这是一款下一代大语言模型，拥有 7700 亿总参数和 490 亿激活参数，上下文窗口超过 100 万 token。它在涵盖编程、办公和科学研究的实际生产力任务中展现出了出色的能力。
Hy4 preview is now available as an open-source model and can also be accessed globally through WorkBuddy and CodeBuddy, as well as Yuanbao, ima and other Tencent products. Users can try the model directly through these applications, or connect to it via API through Tencent Cloud TokenHub and OpenRouter. Upon launch, Hy4 preview will be available for free on WorkBuddy and CodeBuddy for two weeks. Free access to Hy3 on both platforms has also been extended until September 30.
Hy4 预览版现已作为开源模型发布，用户也可以通过 WorkBuddy、CodeBuddy 以及元宝、ima 等腾讯产品在全球范围内访问。用户可以直接通过这些应用程序试用该模型，或通过腾讯云 TokenHub 和 OpenRouter 以 API 形式连接。发布后，Hy4 预览版将在 WorkBuddy 和 CodeBuddy 上免费提供两周。这两个平台上对 Hy3 的免费访问权限也已延长至 9 月 30 日。
Hy4 preview was expanded significantly in model size, context length, and data volume, and the advances in both pre-training and post-training have led to a major leap in overall intelligence, placing the model among the top tier of open-source models. Hunyuan continuously works in deep co-design with products such as CodeBuddy and WorkBuddy, optimizing the real-world user experience across productivity scenarios.
Hy4 预览版在模型规模、上下文长度和数据量上进行了大幅扩展，预训练和后训练的进步带来了整体智能的重大飞跃，使该模型跻身开源模型的第一梯队。混元持续与 CodeBuddy 和 WorkBuddy 等产品进行深度协同设计，优化各种生产力场景下的实际用户体验。
In a blind evaluation conducted internally by Tencent involving 163 experts and 203 engineering tasks, Hy4 preview scored an average of 2.99 out of 4.00, slightly ahead of GLM-5.3 (2.92/4.00) and Kimi K3 (2.94/4.00). Designed for productivity, Hy4 preview was developed using high-quality training data co-created with Tencent experts across software engineering, gaming, finance, security, and other domains, as well as through deep co-design with products such as WorkBuddy. This has helped drive significant improvements across a wide range of real-world productivity tasks.
在腾讯内部进行的盲测中，涉及 163 名专家和 203 项工程任务，Hy4 预览版的平均得分为 2.99 分（满分 4.00 分），略高于 GLM-5.3（2.92/4.00）和 Kimi K3（2.94/4.00）。Hy4 预览版专为生产力而设计，其开发使用了与腾讯在软件工程、游戏、金融、安全等领域的专家共同创建的高质量训练数据，并与 WorkBuddy 等产品进行了深度协同设计。这有助于在广泛的实际生产力任务中实现显著改进。
In software engineering, Hy4 preview delivers stronger understanding, planning, debugging, and validation capabilities for long-context development tasks, while also enhancing the visual quality and interaction experience of front-end development. In office productivity and analytical scenarios, the model demonstrates a significantly stronger understanding of complex working environments and enhanced financial analysis capabilities. It has also been optimized for data analysis and cross-document collaboration, supporting the full workflow from information processing through to the creation of documents, spreadsheets, and presentations.
在软件工程方面，Hy4 预览版为长上下文开发任务提供了更强的理解、规划、调试和验证能力，同时提升了前端开发的视觉质量和交互体验。在办公生产力和分析场景中，该模型展现出对复杂工作环境更强的理解力以及增强的财务分析能力。它还针对数据分析和跨文档协作进行了优化，支持从信息处理到创建文档、电子表格和演示文稿的全流程工作。
In game development, Hy4 preview can generate a playable prototype from a single natural-language request, and work effectively with game engines. Developers can then continue refining complex game projects through multi-turn interactions. In scientific research, Hy4 preview demonstrates stronger capabilities in understanding, reasoning through and solving complex research problems, with notable improvements across areas including AI research and development, molecular dynamics simulation, condensed-matter physics and fundamental mathematics.
在游戏开发方面，Hy4 预览版可以根据单一自然语言请求生成可玩的原型，并能与游戏引擎有效协作。开发者随后可以通过多轮交互继续完善复杂的游戏项目。在科学研究方面，Hy4 预览版在理解、推理和解决复杂研究问题方面展现出更强的能力，在人工智能研发、分子动力学模拟、凝聚态物理和基础数学等领域均有显著提升。
Notably, Hy4 preview also contributed to its own development process, participating for the first time in the automated optimization of training methods, data strategies, evaluation frameworks, and low-level operators. The model proposed approaches, ran experiments, and iterated based on the results, with the resulting code, logs, and feedback feeding into subsequent rounds of exploration. This established an early-stage recursive self-improvement loop.
值得注意的是，Hy4 预览版还参与了自身的开发过程，首次参与了训练方法、数据策略、评估框架和底层算子的自动化优化。该模型提出方案、运行实验并根据结果进行迭代，将产生的代码、日志和反馈输入到后续的探索轮次中。这建立了一个早期的递归自我改进循环。
Hy4 preview has also autonomously analyzed bottlenecks in its inference system and carried out multiple rounds of optimization on areas such as operator fusion and communication optimization. These improvements increased end-to-end throughput by 31.8% compared with the baseline, with consistent gains across different context lengths and concurrency levels. This demonstrates the model’s ability to autonomously optimize its own inference infrastructure.
Hy4 预览版还自主分析了其推理系统中的瓶颈，并在算子融合和通信优化等方面进行了多轮优化。这些改进使端到端吞吐量较基准提升了 31.8%，在不同上下文长度和并发级别下均保持了一致的增益。这证明了该模型能够自主优化其推理基础设施的能力。
Hy4 preview continues to offer cost efficiency, helping make advanced AI more widely accessible. API pricing is set at USD 0.834 per million input tokens, USD 2.501 per million output tokens and USD 0.042 per million tokens for cache hits. Through a preview-first approach, followed by official releases, Hunyuan continuously incorporates real-world feedback into its research and development process, enabling its models to improve by solving real-world problems. The next batch of models in the Hy4 series is expected to roll out soon.
Hy4 预览版继续提供成本效益，助力先进人工智能的普及。API 定价为每百万输入 token 0.834 美元，每百万输出 token 2.501 美元，缓存命中每百万 token 0.042 美元。通过“预览先行、正式发布跟进”的策略，混元持续将实际反馈纳入研发流程，使模型通过解决现实问题不断进步。Hy4 系列的下一批模型预计将很快推出。