It was introduced that compared to the MoE (Mixture of Experts) model structure of Qwen3, Qwen3-Next has made the following core improvements: hybrid attention mechanism, high sparsity MoE structure, ...

On September 12, Alibaba Tongyi released the next-generation foundation model architecture Qwen3-Next and the model based on the new architecture, Qwen3-Next-80B-A3B, which includes two versions: the ...

Alibaba’s Qwen3-Next-80B-A3B-Base model contains 80 billion parameters but activates only 3 billion during inference. The ...

There’s a lot of AI news this week from Google, Microsoft, OpenAI, and Anthropic, which we’ll cover in the news section below. Most of the product innovations these companies are rolling out are built ...