Shanghai AI Lab opens Atria Dawn Preview for long-horizon research agents

ATRIA, from the Shanghai Artificial Intelligence Laboratory, released Atria Dawn Preview — an open agentic model built to carry scientific work from a method in the literature to executable experiments, reproducible metrics, and an inspectable report. The model card and weights are live on Hugging Face; documentation and evaluations are published at atria-asi.ai.

SOFTWARE desk — same-day primary open release of a large Chinese lab agent model aimed at end-to-end research automation, with public weights and reported tool-use/security benchmarks.

Built on a 744-billion-parameter mixture-of-experts foundation the lab names GLM-5.2. Mixture-of-experts, or MoE, means a large model that turns on parts of itself for each query instead of running every parameter every time. The lab says the model is built for continuous tool use, multi-step tasks, and results other people can check. File that foundation size, the MoE picture, and that job as the lab’s. This desk did not count parameters or watch a tool loop.

The same model card positions the work across four workflows: Discovery — turning a research question into an executable experiment plan. Creation — building software, apps, and machine-learning systems. Delivery — turning documents and data into reports and other structured files. Cybersecurity — analyzing security issues and checking fixes in authorized environments. File that four-way map as the lab’s. This desk did not run a Discovery plan or a security check.

Lab-reported scores on the Hugging Face model card — not independent Bad Signal runs: AutomationBench 53.8; CyberGym 86.5; BFCL v4 77.0; DeepSearchQA 96.0; BrowseComp 92.5. AutomationBench is a tool-use automation bench. CyberGym is a security-task bench. BFCL is the Berkeley Function-Calling Leaderboard, a tool-calling test. File those five figures as lab-reported. The same card and the Newsfile note that some individual metrics may change as evaluation continues, and that the project site carries current comparison tables. This desk did not rerun the benches.

Availability, lab: instruct and FP8 weights are on Hugging Face at launch under an MIT license. MIT here means you can use, change, and share the weights, including commercially, if you keep the license notice. The lab also publishes local-deploy notes for SGLang and vLLM and hosted API endpoints by region. File the Hugging Face / MIT / local-or-hosted split as the lab’s. Version pins and install commands stay in Sources. This desk did not download the weights or call the API.

Plain English for the rest of the card: agentic = the model is meant to use tools and keep working across many steps, not just answer one chat turn. MoE = mixture-of-experts, a big model that activates parts of itself per query. open weights = downloadable model parameters you can run yourself. MIT = a permissive open-source license. Preview here is the product name, not a closed beta.

PRIMARY here: the Hugging Face model card and weights for internlm/Atria-Dawn-Preview — Tier A PRIMARY lab source, the original record — plus the same-day Newsfile company distribution. The project site at atria-asi.ai is docs and evaluation-home context, not a second originating newsroom. The arXiv paper 2609.15818 is the technical write-up. The 15 Sep 2026 open release, the 744B MoE GLM-5.2 foundation, the Discovery / Creation / Delivery / Cybersecurity map, the lab-reported AutomationBench / CyberGym / BFCL v4 / DeepSearchQA / BrowseComp figures, the instruct + FP8 Hugging Face weights, the MIT license, and the SGLang / vLLM local-deploy plus hosted-API notes are lab-attributed. Benchmark numbers stay lab-reported — not independently run here. NOT claimed: Western commercial customers, pricing, that this desk reproduced the weather-forecast demo, independently verified benches, that Preview means closed invite-only, a stock tip, or investment advice. Distinct from the already-filed datacebo-sdv-2, testmu-kane-assurance, cyberhill-cerebro-llms, gemini-3-8-live, gensyn-open-1b-auditable, nofire-brig-open-source, hyundai-atria-data-flywheel, and openai-bio-bills-endorsement.