Researchers who want a big open agentic model under MIT can download it, but running 744B parameters locally is heavy and API pricing is unpublished.
Kenning's take
What it does
Atria Dawn Preview is a preview release of an agentic model built on the 744B-parameter GLM-5.2 base. It is aimed at multi-step work such as deep research, coding, building reports and security testing, with a 256K context window. Weights are on Hugging Face and ModelScope under the MIT licence, and there is a hosted API.
- Built on the 744B-parameter MoE GLM-5.2 model
- 256K context window
- FP8 quantised version available
- Open weights under the MIT licence
- Hosted API with international and China endpoints
Who it suits. AI researchers and developers who want an open agentic model to run locally or call through an API.
Key facts
- Category
- AI models
- Made by
- Shanghai Artificial Intelligence Laboratory
- Based in
- China
- Released
- 11 September 2026
- How to get it
- Download the open weights from Hugging Face or ModelScope, or use the hosted API at api.atria-asi.ai
- Works on
- API, Self-hosted (SGLang, vLLM)
- Free plan
- Open weights free under MIT licence
FAQ
When was Atria Dawn Preview released?
The model was published on Hugging Face on 11 September 2026.
Is Atria Dawn Preview open source?
The weights are released under the MIT licence on Hugging Face and ModelScope.
Who made Atria Dawn Preview?
Shanghai Artificial Intelligence Laboratory, building on the GLM-5.2 base model.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.