Qwen3.5-397B-A17B
Alibaba (Qwen)Gated Delta Networks combined with sparse Mixture-of-Experts deliver high-throughput inference with minimal latency and cost overhead.
3 more from this model
The model alternates between Gated DeltaNet layers (linear attention) and full attention layers in roughly a 3:1 ratio.
Hidden Layout: 15 * (3 * (Gated DeltaNet -> MoE) -> 1 * (Gated Attention -> MoE))
Hidden Layout: 10 × (3 × (Gated DeltaNet → MoE) → 1 × (Gated Attention → MoE))