AI Briefing
KO

Tencent Unveils Hy3-preview

·2026.04.23 23:17

Key point

Tencent has unveiled Hy3-preview, an MoE model with 21B active parameters out of a total of 295B parameters.

Details

Tencent has unveiled Hy3-preview, a new open-source MoE (Mixture-of-Experts) model.

This model has a total of 295B parameters, but adopts a structure where only 21B parameters are activated during inference, enhancing computational efficiency.

Key specifications:

  • Model architecture: MoE (Mixture-of-Experts)
  • Total parameters: 295B
  • Active parameters: 21B
llmopen-sourcemoetencenthy3-preview
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

RedditOpen-Weight CLM Released as Jev Alternative, Agent Decision-Making 4x FasterPyTorchKRPostgreSQL Extension pgjev Released, Supporting Natural Language Filtering Without EmbeddingsRedditMica v0.1 4B Released: Open-Source Decision Model for Agent Loops Trained for Under $30PyTorchKRStepFun Releases Step 5 PreviewHacker NewsJevper Released: A Jev Classification Wrapper for OpenAI-Compatible Models with Automatic Logprobs DetectionKakaoKakao Improves Valid Play Rate by 20% with MoE-Based Personalized Ranking Model Using GDCN