Shanghai AI Lab Unveils Open-Source ‘Intern-S1’ Science Multimodal Model

Shanghai, China – At the World Artificial Intelligence Conference (WAIC 2025) in Shanghai, the Shanghai AI Laboratory (Shanghai AI Lab) has launched and open-sourced ‘Intern-S1,’ a groundbreaking scientific multimodal model. This marks a significant leap forward in AI-driven scientific discovery, offering researchers a powerful new tool to tackle complex challenges across various disciplines.

A Scientific All-rounder:

Intern-S1 stands out as the first open-source general-purpose model to integrate specialized scientific capabilities. It excels in cross-modal scientific analysis, adeptly interpreting intricate scientific data formats, including chemical formulas, protein structures, and seismic wave signals. Impressively, it surpasses the performance of leading closed-source models like Grok-4 on benchmarks across chemistry, materials science, and earth science.

Bridging the Gap in Scientific Understanding:

Traditional scientific research often relies on single-modal analysis, which can fall short when exploring complex phenomena, especially in interdisciplinary fields. Intern-S1 addresses this limitation by leveraging the strengths of the Intern large model family. It achieves a high level of balanced performance in both language and multimodal understanding within a single model. By incorporating a wealth of multidisciplinary knowledge and specifically enhancing scientific capabilities, Intern-S1 offers a superior solution compared to existing open-source multimodal models.

Redefining Scientific Productivity:

While large language models have demonstrated remarkable progress in areas like chatbots, image generation, and code creation, the scientific community has been eagerly awaiting an AI partner that truly understands science. Current mainstream models, despite their proficiency in natural language processing and image recognition, often struggle with the complexity, precision, and specialization required for scientific tasks.

Existing open-source models often lack the deep understanding of complex scientific data necessary to meet the stringent demands of scientific research. Closed-source models, while potentially more powerful, often come with high deployment costs and limited controllability. Intern-S1 aims to overcome these hurdles by providing a powerful, accessible, and transparent platform for scientific exploration.

Intern-Discovery: A Platform for Collaborative Scientific Advancement:

Alongside Intern-S1, the Shanghai AI Lab has also launched Intern-Discovery, a scientific discovery platform built upon the new model. This platform aims to enhance the capabilities of researchers, research tools, and research subjects, fostering collaboration and driving scientific research towards a new era of scaling laws.

Availability:

Researchers and developers can access Intern-S1 through the following links:

Experience Page: https://chat.intern-ai.org.cn/
GitHub: https://github.com/InternLM/Intern-S1
HuggingFace: https://huggingface.co/internlm/Intern-S1-FP8
ModelScope: https://modelscope.cn/models/ShanghaiAILaboratory/Intern-S1

Looking Ahead:

The release of Intern-S1 and Intern-Discovery represents a significant step towards democratizing AI-powered scientific research. By providing an open-source, powerful, and versatile tool, the Shanghai AI Lab is empowering researchers worldwide to accelerate discovery and address some of the world’s most pressing challenges. The future of scientific exploration looks brighter with the advent of AI models that can truly understand and contribute to the scientific process.

References: