ScienceOne
| ScienceOne | |
|---|---|
 | |
| Developer | Chinese Academy of Sciences |
| Release | 2025 |
| Website | www |
ScienceOne, also known as Panshi (磐石), is an artificial intelligence platform for scientific research developed by the Chinese Academy of Sciences (CAS) and its Institute of Automation.[1] It can review scientific literature, analyse research data, and select and coordinate specialised software for multistep research tasks.[1][2]
ScienceOne models are used in fields ranging from mathematics and physics to biology, astronomy, materials science and environmental science.[3]
History
[edit]ScienceOne was introduced in April 2025. Its initial tools included S1-Literature, a scientific literature assistant, and S1-ToolChain, which coordinates research software.[1][2] The Panshi scientific foundation model was presented at the World Artificial Intelligence Conference on 26 July 2025.[4] Version 1.5 came out in November 2025 and added tools for evaluating research ideas.[5]
ScienceOne 100 was released in April 2026, bringing the foundation model together with eight field-specific models.[3] In July 2026, CAS released Panshi 2.0.[6]
Features
[edit]ScienceOne combines general-purpose AI models with models designed for scientific tasks. Its S1-Base models are based on Qwen.[1]
S1-Literature is designed for analysing scientific papers, tracing citations and preparing literature reviews.[2]
Panshi 2.0 brings different types of scientific data into a common representation. Reported uses include predicting chemical properties or inferring molecular structures.[6]
Applications
[edit]Reported uses include analysing particle-physics experiments, studying materials, processing astronomical observations and modelling fluid flow.[4][3] According to CAS, the Panshi model system had been deployed across more than 50 of its units by April 2026.[3]
The Shanghai Advanced Research Institute developed ScienceOne–Yuheng, a carbon-accounting system that measures and attributes greenhouse gas emissions.[3][7]
See also
[edit]References
[edit]- 1 2 3 4 강호준; 유재흥 (2025-09-12). 과학을 위한 AI(AI4Science) 연구의 패러다임을 바꾸다 [AI for Science (AI4Science): Transforming the Research Paradigm] (Report) (in Korean). Software Policy & Research Institute. pp. 26–27.
- 1 2 3 "Chinese Institute Launches AI-powered Research Platform "ScienceOne"". Hunan Government Website. Xinhua. 2025-05-07.
- 1 2 3 4 5 孙自法 (2026-04-28). "中国"磐石100"模型体系发布 AI赋能八学科大模型"八仙过海"" [China releases the Panshi 100 model system for eight scientific disciplines]. China News Service (in Chinese).
- 1 2 孙自法 (2025-07-26). "开启科研无限可能 中国团队发布"磐石·科学基础大模型"" [Chinese team releases the Panshi scientific foundation model]. China News Service (in Chinese).
- ↑ 陆成宽 (2025-11-10). ""磐石V1.5：一站式科研平台"发布" [Panshi V1.5: One-stop research platform released]. Science and Technology Daily (in Chinese) – via Chinese Academy of Sciences.
- 1 2 杉村佳織; 澤田莉沙; 阪口幸駿; 田子智久; 永野智己 (2026-09-18). "中国のAI for Science政策、国内各地で具体化 ―各地で進む計画と科学研究基盤の整備―" [China's AI for Science policies take shape across the country: Plans and scientific research infrastructure]. AI for Science Spotlights (in Japanese). Japan Science and Technology Agency.
- ↑ Bela, Victoria (2026-04-22). "How China's new AI carbon accounting model points the finger at the US". South China Morning Post.