Alibaba's Qwen 3.8-Max Activates Only 95B of 2.4T Parameters

Div_pradeep · x · 2026-08-05

Alibaba has reportedly released the Qwen 3.8-Max model. The model features a massive 2.4 trillion total parameters but activates only 95B during inference.

Thanks to this highly efficient Mixture-of-Experts (MoE) architecture, it ranks among the world's best frontier models, drawing attention for its impressive inference efficiency.

Related event: Alibaba Reportedly Launches Qwen 3.8-Max with 2.4T Parameters(2 posts)→

Original post →

More from Models

Models channel →