Atria Dawn officially launches today.
From research questions to verifiable results—now with a 512K context window and stronger performance vs. Preview:
• JobBench: +7.2 pts
• WideSearch: +2.7 pts
• PaperBench: +2.6 pts
Explore:
🌐 Website: atria-asi.ai/
💻 GitHub: github.com/atria-asi/Atria-D…
🚀 Try Atria: discovery.intern-ai.org.cn/t…
🚀 Highly recommend Sage S2.1 Flash, a multimodal model built for iterative scientific research, combining scientific understanding, information retrieval, coding, and tool use.
It supports 10+ scientific input types and ranks first on multiple benchmarks.
🤗 Hugging Face: huggingface.co/internlm/Sage…
🤖 ModelScope: modelscope.cn/models/Shangha…
🧪 Try it: discovery.intern-ai.org.cn/t…
Atria ASI retweeted
🚀 Introducing ToolSpace🌟 — tool-centric environment synthesis for long-horizon agents!
ToolSpace flips data-centric Env scaling🧐:
• Real traces are linked into a tool space, and databases are synthesized only for stateful environments.
• State-aware exploration then turns executable chains into verifiable tasks and checkers.
🔥 Key Results:
• ToolSpace-744B averages 56.6 on 7 long-horizon tool benchmarks, on par with Claude Opus 5
• AutomationBench 54.7 · Harvey LAB 27.5 · SkillsBench 67.8 · OfficeQA Pro 67.1 · JobBench 57.5
• On the new Pro benchmarks, strong models all score below 40
💻 GitHub: github.com/dongguanting/Tool… 📄 Paper: arxiv.org/pdf/2611.05036
🌐 Project Page: toolspace-agent.github.io/
🤗 Models: huggingface.co/collections/d…
📎HF paper: huggingface.co/papers/2611.0…
🙌 Try it, share feedback, and give us a ⭐!
Atria ASI retweeted
Excited to share our latest work 🥳:
🌍 CoworkOS: Scaling Agent Coworkers for Real-world Automation
Building a true “super AI coworker” that handles both complex productivity and long-horizon decision-making.
💡 Three pillars of Design:
1️⃣ Three-layer cowork: Main plans & accepts · Coordinator schedules · Subagents execute (sync/async)
2️⃣ Shared apps: humans & agents co-edit the same docs / CAD / Blender — chat or direct edits
3️⃣ Persistent runtime: hierarchical workspaces + semantic memory + background Jobs → resumable, monitorable
📊 Results:
✅ Competitive with Codex & DeepSeek Harness across 12 productivity benchmarks + 24 real deliverables
✅ RSI-native: 5 rounds of automated post-training lift Qwen3.6-35B-A3B from 33.19 → 42.95 (+9.76) on 8 benchmarks
📄 Paper: github.com/RUC-NLPIR/CoworkO…
🌐 Site: coworkos-ai.github.io/
💻 Code: github.com/RUC-NLPIR/CoworkO…
🎬 Demo: 📷coworkos-ai.github.io/showca…
Atria Dawn officially launches today.
From research questions to verifiable results—now with a 512K context window and stronger performance vs. Preview:
• JobBench: +7.2 pts
• WideSearch: +2.7 pts
• PaperBench: +2.6 pts
Explore:
🌐 Website: atria-asi.ai/
💻 GitHub: github.com/atria-asi/Atria-D…
🚀 Try Atria: discovery.intern-ai.org.cn/t…
Atria ASI retweeted
时隔一个月,上海人工智能实验室AtriaDawn大模型正式版终于上线。对比preview预览版,上下文直接拉到512k,整体能力大幅升级,完全免费可用。不多废话,直接上手体验!
#大模型 #智能体 #人工智能 #ai #科研 #ailab
Atria Dawn officially launches today.
From research questions to verifiable results—now with a 512K context window and stronger performance vs. Preview:
• JobBench: +7.2 pts
• WideSearch: +2.7 pts
• PaperBench: +2.6 pts
Explore:
🌐 Website: atria-asi.ai/
💻 GitHub: github.com/atria-asi/Atria-D…
🚀 Try Atria: discovery.intern-ai.org.cn/t…
Atria ASI retweeted
Introducing Atria Dawn🫡 — the official release!
Atria is a 700B agentic foundation model for long-horizon tasks and real-world productivity 🙌
Upgraded from Preview: 512K context, stronger tool use and ML coding for stable long-horizon execution🥳.
📈 Excited results over Preview:
• Automation Bench 54.7%
• WideSearch 84.8%
• Workspace Bench 66.1%
• MLE Bench 88.2%
From protein structure training & CAD modeling to knowledge viz, PPT/poster design & data analysis
🌐 Homepage: atria-asi.ai
💻 GitHub: github.com/atria-asi/Atria-D…
🤗 Hugging Face: huggingface.co/internlm/Atri…
🚀 Try it: discovery.intern-ai.org.cn/t…
Atria ASI retweeted
Stay tuned for Atria—more RSI explorations and exciting features are coming soon 🥳!
😃“From research questions to verifiable results” is the promise.
😉The proof? It ships.
🚀Meet Atria Dawn Preview @AtriaASI — an agentic foundation model for long-horizon tasks and controlled self-evolution, built on the Intern large model.
🌦️ 1 min → global weather forecast
🖥️ 20 min → OS from scratch
🧠 AI building AI, with verification in the loop Not uncontrolled RSI.
Verifiable. Reproducible. Controllable.
Open preview ↓
🌐 atria-asi.ai
💻 github.com/atria-asi/Atria-D…
🤗 huggingface.co/internlm/Atri…
🤖 modelscope.cn/models/Shangha…
🐦 teapaw.org/AtriaASI
#AtriaDawn #AgenticAI #RSI #OpenSource
Atria ASI retweeted
😃“From research questions to verifiable results” is the promise.
😉The proof? It ships.
🚀Meet Atria Dawn Preview @AtriaASI — an agentic foundation model for long-horizon tasks and controlled self-evolution, built on the Intern large model.
🌦️ 1 min → global weather forecast
🖥️ 20 min → OS from scratch
🧠 AI building AI, with verification in the loop Not uncontrolled RSI.
Verifiable. Reproducible. Controllable.
Open preview ↓
🌐 atria-asi.ai
💻 github.com/atria-asi/Atria-D…
🤗 huggingface.co/internlm/Atri…
🤖 modelscope.cn/models/Shangha…
🐦 teapaw.org/AtriaASI
#AtriaDawn #AgenticAI #RSI #OpenSource
Introducing Atria Dawn Preview:
From Research Questions to Verifiable Results.
Atria Dawn Preview is an agentic foundation model built for long-horizon tasks, helping researchers and engineers turn open-ended questions into executable, verifiable, and reproducible outcomes.
Explore Atria Dawn Preview:
🌐 Website: atria-asi.ai/
💻 GitHub: github.com/atria-asi/Atria-D…
🤗 Hugging Face: huggingface.co/internlm/Atri…
🤖 ModelScope: modelscope.cn/models/Shangha…
🚀 Try Atria:
EN:api.atria-asi.ai/
ZH:discovery.intern-ai.org.cn/t…
If you have any questions, feedback, or want to share what you’ve made, feel free to join our Discord: discord.gg/r35zM7b8u
2
802
We’re excited to highlight SETA, an open-source RL environment for training and evaluating AI agents from the @CamelAIOrg .
A valuable resource for researchers and developers exploring agentic reinforcement learning and complex task environments.
Check it out and support the CAMEL-AI community! 🚀
🔗 github.com/camel-ai/seta
Atria ASI retweeted
I'm also excited to share our team's study on human–AI collaboration. Throughout the training process, we (almost) achieved recursive self-improvement (RSI)— and I believe fully automated RSI is not far away🥳 .
From solving tasks to partnering on entire projects.
We introduce Atria Dawn: The Dawn of Agentic Superintelligence — our study of what happens when agents move beyond answering questions to exploring, building, testing, and producing verifiable outcomes.
🔹 Verifiable Experience Pipeline
🔹 16 real-world benchmarks
🔹 #1 on 5 benchmarks
🔹 769 real task records from 56 participants
🔹 A case study of human–AI collaboration
The goal isn’t just more capable agents. It’s agents that can contribute to real research and engineering workflows under human oversight.
📄 Paper: huggingface.co/papers/2609.1…
Atria ASI retweeted
Meet Atria Dawn Preview, a new AI model from the Atria team, a collaboration between Shanghai AI Laboratory and multiple universities and research institutions.
Built for scientific productivity, Atria Dawn Preview is designed to tackle complex research tasks continuously, turning long-horizon work into progressively verifiable outcomes.
Its focus goes beyond solving individual tasks: Atria aims to strengthen end-to-end delivery in real-world workflows, from scientific automation to everyday office work.
Website: atria-asi.ai/
We’re grateful to OpenBMB for supporting Atria Dawn Preview with UltraData-SFT-Agent-2609.
The dataset’s high-quality agent trajectories have been valuable in developing Atria’s capabilities on agentic tasks.
🔗 huggingface.co/datasets/open…
From solving tasks to partnering on entire projects.
We introduce Atria Dawn: The Dawn of Agentic Superintelligence — our study of what happens when agents move beyond answering questions to exploring, building, testing, and producing verifiable outcomes.
🔹 Verifiable Experience Pipeline
🔹 16 real-world benchmarks
🔹 #1 on 5 benchmarks
🔹 769 real task records from 56 participants
🔹 A case study of human–AI collaboration
The goal isn’t just more capable agents. It’s agents that can contribute to real research and engineering workflows under human oversight.
📄 Paper: huggingface.co/papers/2609.1…
Cybersecurity
Cybersecurity tasks put long-horizon agent capabilities to the test. In an authorized, isolated environment, Atria Dawn Preview combines environment understanding, vulnerability analysis, tool use, and result verification across the entire task.
From analyzing an application and identifying its attack surface to validating vulnerabilities, executing exploits, and applying fixes, Atria carries the task through multiple stages with continuous verification.
🎥 See Atria Dawn Preview in action.

















