Tencent has launched the Hy3 model with 295 billion parameters under an Apache 2.0 license, removing previous deployment restrictions. This change aims to enhance accessibility for enterprises, particularly in regions previously excluded from similar technologies.
Tencent's Hunyuan team released Hy3, a 295-billion-parameter model, under the permissive Apache 2.0 license. This shift addresses previous licensing issues that limited deployment in the EU, UK, and South Korea, which had hindered enterprises interested in utilizing open-weight models.
The full version of Hy3 has evolved rapidly, going from a preview release in April to a fully functional model in just ten weeks. This acceleration involved feedback from over 50 internal teams, which allowed Tencent to refine the model based on practical testing and interaction.
Hy3 maintains an architectural layout with 295 billion total parameters and leverages a Mixture-of-Experts mechanism. It offers 21 billion active parameters for each forward pass, utilizing a top-8 routing across 192 experts, along with a large context window of 256K. Tencent asserts that Hy3 outperforms other models of comparable size, promising greater reliability for production use.
The release and licensing change of Hy3 positions Tencent as a significant player in the open-source AI model landscape, especially as it becomes accessible to a broader set of enterprises. The competitive advantage stems not only from its technical specifications but also its openness, fostering adoption within markets that have previously faced legal restrictions.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
Tencent has launched the Hy3 model with 295 billion parameters under an Apache 2.0 license, removing previous deployment restrictions. This change aims to enhance accessibility for enterprises, particularly in regions previously excluded from similar technologies.