PyyanThe AI Index

Pyyan / Compare / DeepSeek V4 Flash Vision Exp vs Atria Dawn Preview

DeepSeek V4 Flash Vision Exp vs Atria Dawn Preview

2 of 5

Open-Weight Models · verified 16 Sept 2026

DeepSeek V4 Flash Vision ExpDeepSeekdeprecated
SAtria Dawn PreviewShanghai AI Laboratorycurrent
3 slots left
SpecificationDeepSeek V4 Flash Vision ExpAtria Dawn Preview
SummaryExperimental, says so in the name, and shipped anyway.744B agentic weights under the MIT licence, published without a launch.
Context128K256K
Input ($/Mtok)$0.44Not published
Output ($/Mtok)$1.32Not published
LicenceOpen weightsMIT
Max output32KNot published
ParametersNot published744B total
Released21 Aug 202614 Sep 2026
Model IDdeepseek-ai/DeepSeek-V4-Flash-Vision-Expinternlm/Atria-Dawn-Preview
CategoryOpen-Weight ModelsOpen-Weight Models
Official—Shanghai AI Laboratory ↗

Highlighted rows are where these differ.

DeepSeek V4 Flash Vision Exp

  • Retired 10 September 2026: requests now route to DeepSeek V4.1 Flash
  • Was DeepSeek's first Flash tier with image input

Best for cheap multimodal trials.

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 →