Tencent has released Hy3, a 295B-parameter open-source MoE model under Apache 2.0. With only 21B active parameters, it outperforms models 5X its active size and reduces hallucination rates to 5.4%.

Tencent Releases Hy3: A Game-Changing Open Source Model

Tencent has officially released its AI model Hy3 under Apache 2.0 license. The model uses a Mixture-of-Experts (MoE) architecture with 295 billion total parameters. Of those, 21 billion are active at any given time, plus 3.8 billion parameters for an added MTP layer. It handles context lengths up to 256,000 tokens.

Benchmark Results

Tencent says Hy3 matches the performance of models two to five times its size. In a blind evaluation by 270 experts, Hy3 scored 2.67 out of 4, beating GLM-5.1 at 2.51. Internal testing showed the hallucination rate dropped from 12.5 percent to 5.4 percent — a remarkable 56% reduction.

Technical Specifications

The model is available under Apache 2.0 license on Hugging Face, ModelScope, and GitHub. An FP8-quantized version is also available. Support for platforms like OpenRouter and Cline is planned.

Integrated Products

Tencent has already built Hy3 into its own products, including WorkBuddy, Yuanbao, WeChat, and the game assistant for Path of Exile: Advent. This demonstrates the model's real-world readiness.

Significance for AI Industry

The release of Hy3 shows that Chinese AI companies continue to push open-source AI forward. Models like Hy3 are helping to democratize AI by making powerful capabilities freely available.

What This Means for Developers

For developers, Hy3 represents a new option in the open-source AI stack. Its MoE architecture means it runs efficiently despite its massive size, making it accessible to organizations with moderate computing resources.

Conclusion

Tencent's Hy3 is a significant contribution to open-source AI that will likely accelerate development across the ecosystem.

Tencent Releases Hy3: A 295B-Parameter Open-Source MoE Model That Competes With Models 5X Its Active Size