
Intern-S1-Pro
Trillion-parameter open scientific multimodal model for AI-for-science.
About
Intern-S1-Pro is the Shanghai Artificial Intelligence Laboratory's 1-trillion-parameter scientific multimodal foundation model, open-sourced on February 4, 2026. Built on the lab's SAGE general-specialized fusion architecture, it uses a 512-expert mixture-of-experts design that activates only 8 experts, about 22B parameters, per inference.
The lab positions it as the largest open scientific multimodal model in the community, with headline claims of Olympiad-gold-level math and physics reasoning and top-tier agent capabilities on real research workflows. Architectural additions include Fourier positional encoding and a redesigned temporal encoder aimed at giving the model physical intuition across scales.
Performance and data-sovereignty claims are vendor-reported and should be treated as such; independently reproduced benchmarks are the right check before load-bearing use. As with other China-origin open releases, confirm the exact weight license on the model card before commercial deployment.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 104; Trending score 84/100 from researcher assessment.