PyyanThe AI Index

Pyyan / Compare / Qwen3.8-Flash-Next vs Atria Dawn Preview

Qwen3.8-Flash-Next vs Atria Dawn Preview

2 of 5

Open-Weight Models · verified 16 Sept 2026

Qwen3.8-Flash-NextAlibabacurrent
SAtria Dawn PreviewShanghai AI Laboratorycurrent
3 slots left
SpecificationQwen3.8-Flash-NextAtria Dawn Preview
SummaryThe most downloaded thing on Hugging Face right now, and a stated Qwen4 preview.744B agentic weights under the MIT licence, published without a launch.
Context262K native, ~1M extended256K
Input ($/Mtok)$0.15Not published
Output ($/Mtok)$0.47Not published
LicenceOpen weightsMIT
Max output32KNot published
Parameters125B total, 6B active744B total
Released26 Aug 202614 Sep 2026
Model IDQwen/Qwen3.8-Flash-Nextinternlm/Atria-Dawn-Preview
CategoryOpen-Weight ModelsOpen-Weight Models
Official—Shanghai AI Laboratory ↗

Highlighted rows are where these differ.

Qwen3.8-Flash-Next

  • 125B total parameters with only 6B active per token
  • Alibaba describes it as a preview of the Qwen4 architecture
  • Top of the Hugging Face trending list, with a GGUF build close behind it

Best for open weights at speed.

Full spec sheet →

Atria Dawn Preview

  • Built on Zhipu's open GLM-5.2 rather than trained from scratch
  • 96.0 on DeepSearchQA and 86.5 on CyberGym, both reported by the lab
  • BF16 and FP8 checkpoints, served through SGLang or vLLM

Best for agentic work on your own cluster.

Full spec sheet →