Atria Dawn (full): Shanghai AI Lab's 744B agentic MoE graduates from preview
Shanghai AI Laboratory released the full version of Atria Dawn, its 744B-parameter agentic mixture-of-experts model built on Zhipu's GLM-5.2, doubling context from 256K to 512K tokens and lifting agent-benchmark scores (JobBench 57.5, WideSearch 84.6, PaperBench 89.6).
Why it matters
A state-backed lab's full agentic foundation model with MIT-licensed code — longer context and better research-agent scores than the September preview, with full weights listed as 'coming soon' on Hugging Face and ModelScope.
What you could build with it
A small research-tools startup could wire Atria Dawn into a literature-review agent: given a research question and a corpus of papers, it would produce an executable experiment plan with reproducible metrics and an inspectable report, replacing weeks of manual research-assistant work.
Does it hold up?
Too early to judge for the full version: its weights are still listed as 'coming soon' on Hugging Face and ModelScope, so no independent deployments exist yet. The September preview saw immediate aggregator API adoption and a 140-author technical report, but the full version's larger 512K window is unproven outside the lab's own benchmarks.
Built with Atria Dawn (full)
- Shanghai AI Lab Shipped an Agent Model With No Announcement (dev.to analysis)dev.to · Community writeup analyzing Atria Dawn Preview's weights-first release strategy and its agent-focused design (verifiable execution over chat).
- Shanghai AI Lab Quietly Releases Atria, a Free 744 Billion Parameter AI Agent (Startup Fortune)startupfortune.com · Independent report on the Sept 11 preview release: 744B MoE post-trained on GLM-5.2 for long research and engineering loops, MIT license, hosted API.
- Atria Dawn Preview model hub page (HokAI)hokai.io · Model tracker entry documenting the preview's 256K context, release timeline, and 59.6% SWE-bench Pro 2026 score.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.