Tencent open-sources Hy4 preview with 770B parameters and a 1M-token context
Tencent Hunyuan has open-sourced the Hy4 preview, a new large language model with 770 billion total parameters and a context window exceeding 1 million tokens. The model, released on August 28, features 49 billion activated parameters. It is accessible through Tencent's WorkBuddy and CodeBuddy platforms, as well as Yuanbao and ima, with API access provided via Tencent Cloud TokenHub and OpenRouter. Hy4 preview is designed to enhance productivity across tasks such as coding, office work, data analysis, game development, and scientific research. Tencent is offering free two-week access to WorkBuddy and CodeBuddy users, while API access is priced at $0.834 per million input tokens and $2.501 per million output tokens.
Tencent’s open-sourcing of Hy4 preview, a large language model with 770 billion parameters and a 1 million-token context window, marks a significant move in the competitive Asian AI landscape. The model’s performance in internal blind evaluations, where it scored 2.99 out of 4 against competitors like GLM 5.3 and Kimi K3, suggests a strong contender for enterprise applications. Its availability through WorkBuddy and CodeBuddy, coupled with API access, positions it for broad adoption in productivity tasks from coding to scientific research. The real story for Asia is Tencent’s aggressive push to democratize access to advanced AI, challenging both domestic and international rivals. While the model’s internal performance metrics are promising, the crucial test will be its real-world adoption and integration by developers and businesses across the region. The pricing structure for API access at $0.834 per million input tokens and $2.501 per million output tokens will also influence its market penetration, particularly for smaller startups and developers in Southeast Asia and other emerging markets where cost sensitivity is higher. This release reflects a broader trend of Chinese tech giants investing heavily in foundational AI models and making them accessible to foster ecosystem growth. Tencent’s focus on productivity tasks and its own internal system optimization, which saw a 31.8% increase in end-to-end throughput, indicates a strategic effort to enhance its own operational efficiency while also providing powerful tools to the wider developer community.
