Qwen · Model release
Qwen3.8-Flash-Next: an early look at the Qwen 4 architecture
Alibaba shipped an experimental preview of the architecture that will underpin Qwen 4: 125B total / 6B active MoE plus a 51B n-gram embedding table, at very low active-parameter cost.
What changed
- The vendor-described n-gram component injects a 20-million-entry bigram/trigram table at an early layer — an unusual design aimed at cheap long-context decoding (not yet independently reviewed).
- Native context is 262K, extensible to 1M; input accepts text, image, and video.
- The checkpoint uses the Qwen Community License 1.0, not a standard MIT/Apache grant: commercial Model-as-a-Service and AI Work Assistant businesses need a separate Qwen licence (except qualifying internal use). A commercial product or service above 100 million monthly active users or US$20 million monthly revenue must prominently display the model name. Read the license itself before deployment.
Best fit
- Early architecture testing
- High-throughput, low-active-parameter serving
- Evaluating the Qwen 4 direction
What to be careful about
- It is an experimental preview: expect breaking changes before the Qwen 4 family lands.
- Its public model card reports benchmark results using vendor methodology; these are not independent replications or directly comparable to unrelated harnesses.
- The open checkpoint and hosted Qwen3.8-Flash service are separate products; no hosted price is asserted in this guide.
Dataset snapshot: Qwen 3.8 Flash Next is also represented in the checked comparison dataset. Prices and benchmark figures carry their own field dates.