Chinese Academy of Sciences Unveils ScienceOne 100 AI Model System for Scientific Research
Over 2,000 research tools across 100+ scenarios deployed in 50 institutes.
The Chinese Academy of Sciences (CAS) on Tuesday launched ScienceOne 100, an AI model system designed to accelerate scientific research across multiple disciplines. Built on the ScienceOne foundation model released in 2025, the system integrates eight domain-specific large models covering mathematics, physics, materials science, astronomy, environmental science, aerospace, geosciences, and biology. ScienceOne 100 provides three core functions: Literature Compass for automated literature review and analysis, Innovation Evaluation for tracking cutting-edge trends, and Agent Factory—a platform with over 2,000 research tools supporting autonomous closed-loop workflows.
The system has already been deployed across more than 50 CAS institutes, covering over 100 research scenarios. Its latest version achieves state-of-the-art results on benchmarks for scientific image understanding and manipulation, as well as agentic long-horizon reasoning. Practical applications include high-speed rail flow field reconstruction, spectral identification, materials discovery, adjuvant design, astronomical observation, and ecological research on the Qinghai-Tibet Plateau. This marks a shift from fragmented, isolated AI research toward collaborative, platform-based scientific innovation.
- ScienceOne 100 includes 8 domain-specific models for math, physics, materials, astronomy, environmental science, aerospace, geosciences, and biology.
- Agent Factory provides over 2,000 research tools across 10 research domains with autonomous closed-loop capabilities.
- Deployed in 50+ CAS institutes across 100+ scenarios, achieving SOTA on scientific image understanding benchmarks.
Why It Matters
ScienceOne 100 could democratize AI-assisted research across multiple sciences, accelerating discovery in materials, physics, and biology.