Qwen3 Next 80B A3B

Qwen3 Next 80B with 3B active parameters. Hybrid architecture combining Gated DeltaNet (linear attention) and Gated Attention (standard GQA) with high-sparsity MoE.

Architecture Overview

Loading visualization...

Architecture Summary

27

Total Modules

10

Blocks

17

Kernels

7

Traced Kernels