On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability research paper by Alibaba (Qwen), 2026

Alibaba (Qwen) · Aug 31, 2026 · Architectures · 6 citations · 62 upvotes

Qwen3.8-Flash-Next: a 125B mixture-of-experts model with 6B active that nearly matches its 397B predecessor at 1/9 the training compute.

Summarised by hand from the abstract.

Topic

Date	Paper	Lab	Citations
Jun 12, 2017	Attention Is All You NeedIntroduced the Transformer, the attention-only architecture behind nearly every large language model since.Google · Jun 2017Google	Google	194k
Dec 1, 2023	Mamba: Linear-Time Sequence Modeling with Selective State SpacesA selective state space model that scales linearly with sequence length and matches Transformers on language.Carnegie Mellon University · Dec 2023Carnegie Mellon University	Carnegie Mellon University	9,182

About this paper

Authors: Zihan Qiu, Zekun Wang, Xiao Li and 33 more
arXiv: 2608.30320 · PDF
Citations: 6, 2 influential · Semantic Scholar
Upvotes: 62 · Hugging Face
Lab: Alibaba (Qwen) · on Companies · on Quarterly

Changes

Date	What changed
Sep 24, 2026	New paperAdded to the listSep 24, 2026

Sources: each lab's own papers and arXiv, with citation and upvote counts from Semantic Scholar and Hugging Face. One-line summaries are for orientation, not a substitute for the paper. Logos via logo.dev; trademarks belong to their owners.

What it shows

Other architectures papers

Changes

New papers by email