Shanghai AI Laboratory open-sources Intern-S1-Pro, a 1-trillion-parameter scientific model

Shanghai AI Laboratory open-sourced Intern-S1-Pro, a mixture-of-experts model with 1 trillion total parameters of which only 22 billion, spread across 8 of 512 experts, activate for any given query. The lab positioned it as a foundation for “AI for Science,” aimed at moving AI in scientific domains — chemistry, materials science, life sciences, earth science and physics — from a tool used alongside research toward a system that can help drive discovery directly.

The lab said the model incorporates a Fourier position-embedding technique intended to let it handle signals across very different scales within a single framework, from micro-scale biological signals to macro-scale cosmic fluctuations, and that it reaches gold-medal level on Olympiad-style mathematical and logical reasoning problems, alongside leading results on difficulty-weighted interdisciplinary AI4S evaluations. The model was released with weights on GitHub, Hugging Face and ModelScope, along with an online demo, under terms Shanghai AI Lab said permit free commercial use.

The release is one of several trillion-parameter-class open-weight models to come out of Chinese labs in the AI4S niche, part of a broader pattern in which Chinese state-affiliated and university-linked labs have published large open-weight models more readily than their leading US counterparts, whose frontier models have remained closed since GPT-4.