Skip to content

Qwen3.5-397B-A17B

Released 2026-02 · hybrid · 256K tokens · 4 benchmarks

Editorial notes

Lanzado 16 feb 2026. MoE 397B total / 17B activos. Flagship MoE de la serie 3.5. Hasta 19x decoding speedup a 256K vs Qwen3-Max. Tau2-Bench 86.7 (segundo solo Claude Opus 4.6). Reemplazado al mes por la linea 3.6 pero sigue siendo referencia open-weight. Apache 2.0.

Spec sheet

Company
Alibaba
Country
CN
Type
hybrid
Release
2026-02
Context
256K tokens
Slug
qwen3-5-397b-a17b

Benchmarks (4)

Cite this model
BibTeX · APA

BibTeX

@misc{frontier-qwen3-5-397b-a17b,
  title  = {Qwen3.5-397B-A17B},
  author = {{Alibaba}},
  year   = {2026},
  note   = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
  url    = {https://frontierbenchmarks.com/models/qwen3-5-397b-a17b}
}

APA

Alibaba (2026). Qwen3.5-397B-A17B [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/qwen3-5-397b-a17b

Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.