295 billion parameters but only 21 billion activated, this sparsity design is quite interesting, OpenRouter is already playable, let me test it first.

View Original
CoinNetwork
CoinWorld News, Tencent has released the latest flagship model of the Hunyuan series, Hy3 preview, adopting a MoE (Mixture of Experts, where multiple small expert networks are activated on demand) architecture, with a total of 295 billion parameters and 21 billion activated parameters. The model has been launched on the OpenRouter API and Hugging Face demo, but the open-source weights have not yet been released. This is the first flagship model led by former OpenAI researcher and Tencent Chief AI Scientist Yao Shunyu after he joined, and it is also the output of Tencent's rebuilding of its foundational model R&D system in February of this year, produced in less than three months. Tencent positions it as "on par with top domestic models."
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned