Tencent bets on 295B Hy3 MoE model with a focus on agentic skills

emmanuelvivier · x · 2026-08-04

Tencent is developing a new Mixture-of-Experts (MoE) model named Hy3. With a massive 295 billion parameters, the company is shifting focus away from merely scaling model size to prioritizing robust agentic capabilities, aiming to handle complex, real-world agentic tasks.

Original post →

More from Models

Models channel →