AI Trends

Tencent Officially Open-Sources Hy4, Making the Entire Hunyuan LLM Series Open Source

IT Home · Artificial IntelligenceUpdated Nov 20 22:06

AI summary

Tencent’s Hunyuan team has officially open-sourced Hy4, with the model weights now available on GitHub, Hugging Face, ModelScope, and GitCode, making the entire Hunyuan LLM series open source. Hy4 uses a mixture-of-experts (MoE) architecture with 770B total parameters, 49B active parameters, and a context window exceeding 1 million tokens. Hy4 and Hy4 Preview are licensed under Apache 2.0, allowing commercial use and modification. A System Card covering cybersecurity and biosafety evaluations has also been released.

Why read this

SelectedAI score 77

The material lists Hy4’s total and active parameter counts, context window, and license, facilitating the assessment of local deployment and commercial-use conditions.

Topics

Tags

Body · AI translation

Tencent Officially Open-Sources Hy4, Fully Opening Up the Hunyuan LLM Series

2026-11-20 11:47 ITHome - Wenzhou

ITHome reported on November 20 that Tencent's Hunyuan team today announced the official open-source release of the Hy4 large model. The model weights are now available simultaneously on GitHub, Hugging Face, ModelScope, and GitCode, allowing developers to download and deploy them independently.

With this release, the entire Hunyuan LLM series has now been open-sourced, covering Hy3 preview, Hy3, Hy4 preview, Hy4, Hy4 flash, and other versions.

The official version of Hy4 was first released on November 4. This open-source release provides the model weights previously announced. Hy4 uses a mixture-of-experts (MoE) architecture, with 770B total parameters, 49B activated parameters, and a context window exceeding 1 million tokens.

As ITHome previously reported, the cost-effective Hy4 flash was made available alongside the flagship Hy4. The Hy4 flash model has 224B total parameters, including 64B Engram parameters, activates 12B parameters per token, and supports multimodal input and a context length of 1 million tokens. Its API pricing is substantially lower than Hy4's.

Regarding open-source licensing, Hy4 and Hy4 preview use the Apache 2.0 license, which permits commercial use and modification. Tencent Hunyuan also released the Hy4 model System Card, disclosing evaluation results in areas including cybersecurity and biosafety.

According to the official statement, Hy4's training data includes only defensive cybersecurity content, such as PoC generation and vulnerability remediation, while offensive data involving vulnerability exploitation, reverse engineering, and penetration was excluded. Audits by Tencent's Xuanwu Lab and Keen Security Lab found no evident risk of offensive capabilities.

References

Source

IT Home · Artificial Intelligence