Reportedly: Shanghai AI Laboratory Releases Atria Dawn Full Version With a 512K Context Window
Reported by Pandaily · not yet confirmed by the company or a second independent outlet. We update this page when it is.
Shanghai AI Laboratory released the full version of its Atria Dawn agent model, built on a 744B MoE GLM-5.2 base, with a 512K context window. The release improves scores on JobBench, WideSearch, and PaperBench over the September preview. Full weights are still marked 'Coming soon', Pandaily reported.
China context
- Original name
- 上海人工智能实验室
- Outside China
- Not stated in the sources yet
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Builders outside China cannot yet access the weights, so they cannot test or integrate the model. The next signal is the weights becoming available on a public repository.
- For investors
- Investors should note that the model's benchmark improvements are company-reported and not independently verified. The pending weights release is a key milestone to watch.
Shanghai AI Laboratory has released the full version of Atria Dawn, an agent model built on a 744B MoE GLM-5.2 base with a 512K context window. The release moves past the September preview with higher scores on JobBench, WideSearch, and PaperBench. Full weights are still marked 'Coming soon'.
Atria Dawn is built on a 744B mixture-of-experts (MoE) GLM-5.2 base and supports a 512K context window. The full version improves scores on JobBench, WideSearch, and PaperBench compared to the September preview. Full weights are not yet available.
Developers outside China cannot yet access the weights, so they cannot evaluate or deploy the model; the 'Coming soon' status means the open-weights release is still pending. The next verifiable signal is the publication of the weights on a platform like Hugging Face.
The model's 512K context window and improved agent benchmarks could make it useful for long-document processing and complex agent tasks, but without weights, businesses cannot adopt it yet.
The next observable signal is the release of the full weights, which would allow independent evaluation and deployment. Until then, the model's capabilities are known only from Shanghai AI Laboratory's reported benchmark scores.