Introducing Hy4 Preview
Tencent releases Hy4, a 770B open-weight LLM with 49B active params and 1M token context, 2.6× larger than Hy3.
·
Archive
The Archive
Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
tencent/Hy3
Tencent releases Hy3, a 295B-param MoE model with 21B active params under Apache 2.0, claiming performance parity with 2-5x larger open-source competitors.
·
The frontier reasoning race is starting to look like a crowded subway station
We went from chasing GPT4 to looking at graphs with GPT5.4 xhigh, Gemini 3.1Pro, and now Hy3 preview completely shaking up the leaderboard. Look at that CHSBO 2025 chart Hy3 preview scoring 87.8 over Gemini and GPT. What a time to be alive, but honestly, my brain can't keep up with the version numbers anymore. What's your take? Is Hy3 actually punching at this level in real-world coding/math, or is it just benchmark hardening?
··
Tencent Releases Hy3 preview - Open Source 295B 21B Active MoE
Tencent releases Hy3 preview: 295B MoE model with 21B active parameters on Hugging Face.
··
← Front Page4 matches