China introduced a powerful new tool to advance scientific discovery across many disciplines. Known as the ScienceOne AI model, this innovation supports faster, smarter, and more accurate research. The Chinese Academy of Sciences revealed the system during a major artificial intelligence event, presenting it as a key part of China’s research future.
Right from the beginning, developers built the ScienceOne AI model to understand complex scientific data. It processes waveforms, spectra, and scientific fields with advanced capabilities. In addition, the system enables literature reading, knowledge reasoning, and automatic coordination of computing tools.
Twelve research institutes in China worked together to build the ScienceOne AI model. These include leaders in automation, computing, mathematics, physics, and high-energy science. By combining efforts, they created a system that works across disciplines with great efficiency.
Drawing strength from China’s open-source AI base, the model adds deep scientific customization. Powerful domain tools such as AlphaFold and MatterGen further enhance its capabilities. As a result, the system functions seamlessly in real-world scientific environments.
Researchers can apply the model at different stages of their work. For instance, it assists with reading articles, forming research questions, and selecting the right tools. Scientists no longer need to switch between platforms or struggle with data silos.
Moreover, evaluations confirm that the ScienceOne AI model performs strongly across disciplines. It demonstrates mastery in math, physics, chemistry, astronomy, earth science, and biology. The model also excels in complex reasoning and scientific tool use.
To expand its impact, the team developed two AI agents based on the ScienceOne framework. The first agent helps researchers read and analyze academic papers more efficiently. By tapping into over 170 million articles, it reduces review time from days to minutes.
Meanwhile, the second agent improves planning by assisting with tool selection and task execution. With access to over 300 computing tools, it manages workflows and lets scientists stay focused on research goals.
Already, the ScienceOne AI model powers several real-world science platforms. In biology, it speeds up target identification. For physics, it enhances particle simulations. In chemistry, it sharpens molecule predictions. It also coordinates telescope networks and supports structural innovation for high-speed trains.
Altogether, the model signals a major step forward for AI in science. With ongoing improvements, China aims to make research more intelligent, efficient, and accessible.
