NextFin

Shanghai AI Lab Releases and Open-Sources the First Multimodal Foundation Model with Scientific Reasoning

Summarized by NextFin AI
  • Shanghai AI Laboratory launched and open-sourced Intern-S1, a scientific multimodal foundation model at the 2025 World Artificial Intelligence Conference.
  • Intern-S1 integrates advanced scientific capabilities and features a cross-modal scientific reasoning engine for problem-solving in various scientific disciplines.
  • In benchmark tests, Intern-S1 outperformed leading closed-source models like Grok-4 on specialized scientific tasks, demonstrating its superior capabilities.
  • The lab also unveiled Intern-Discovery, a platform for researchers to support the full cycle of hypothesis generation to experimental validation, now open for global trial access.

AsianFin — At the Frontier of Science plenary session of the 2025 World Artificial Intelligence Conference, Shanghai AI Laboratory officially launched and open-sourced Intern-S1, a scientific multimodal foundation model developed under its “Intern” model family.

Intern-S1 is the first open-source general-purpose model to integrate advanced scientific capabilities. Building on the existing Intern series, it pioneers a “cross-modal scientific reasoning engine” that enables deep understanding and problem-solving across disciplines such as chemistry, materials science, and earth sciences. In benchmark tests, Intern-S1 has outperformed leading closed-source models like Grok-4 on specialized scientific tasks.

Shanghai AI Lab reports that Intern-S1 also leads mainstream open-source multimodal models—such as InternVL3 and Qwen2.5-VL—in overall cross-modal capabilities.

In parallel, the lab unveiled Intern-Discovery, a scientific discovery platform designed to support researchers through the full cycle of hypothesis generation to experimental validation. The platform is now open to global applicants for trial access.

Explore more exclusive insights at nextfin.ai.

Insights

What is the concept of a multimodal foundation model?

How does the Intern-S1 model differ from traditional AI models?

What are the key features of the cross-modal scientific reasoning engine?

What were the benchmark results of Intern-S1 compared to Grok-4?

How has user feedback been regarding the Intern-S1 model?

What trends are emerging in the field of multimodal AI models?

What are the recent developments announced at the 2025 World Artificial Intelligence Conference?

How does the open-sourcing of Intern-S1 impact the AI community?

What potential challenges could arise from the use of Intern-S1 in scientific research?

What controversies exist surrounding the use of AI in scientific reasoning?

How does Intern-S1 compare to other open-source models like InternVL3 and Qwen2.5-VL?

What historical examples exist of significant advancements in AI modeling?

In what ways could the Intern-Discovery platform transform scientific research?

What implications does the launch of Intern-S1 have for future AI developments?

How might the integration of advanced scientific capabilities in AI evolve in the coming years?

What limitations does the Intern-S1 model currently face?

What specific fields of science can benefit the most from the Intern-S1 model?

How do researchers plan to validate the hypotheses generated by the Intern-Discovery platform?

What are the long-term impacts of open-sourcing AI models like Intern-S1?

What ethical considerations arise with the deployment of advanced AI models in research?

Search
NextFinNextFin
NextFin.Al
No Noise, only Signal.
Open App