Tencent releases Hy3 open-source model that allegedly matches models up to five times its active size
Quick Answer
Tencent has launched the Hy3 AI model, featuring a Mixture-of-Experts architecture with 295 billion parameters, claiming performance on par with models two to five times its size.
Quick Take
Tencent has launched the Hy3 AI model, featuring a Mixture-of-Experts architecture with 295 billion parameters, claiming performance on par with models two to five times its size. In evaluations, Hy3 scored 2.67 out of 4, outperforming GLM-5.1, and reduced hallucination rates from 12.5% to 5.4%. The model is available under an Apache 2.0 license on multiple platforms.
Key Points
- Hy3 utilizes a Mixture-of-Experts architecture with 295 billion total parameters.
- The model is claimed to match performance of models 2-5 times its size.
- Hy3 scored 2.67 in expert evaluations, surpassing GLM-5.1's score of 2.51.
- Hallucination rates decreased from 12.5% to 5.4% in internal testing.
- Hy3 is available on Hugging Face, ModelScope, and GitHub under Apache 2.0.
📖 Reader Mode
~1 min readTencent has officially released its AI model Hy3. 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. 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.
Hy3 is available under an 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. Tencent has already built the model into its own products, including WorkBuddy, Yuanbao, WeChat, and the game assistant for "Path of Exile: Advent."
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
