The rapid advancement of Artificial Intelligence has been propelled by a dual-track ecosystem of proprietary and open-source models. However, for the university research community, the rise of "open weights" and fully open-source architectures represents more than just a resource — it is a fundamental necessity for scientific rigour. This talk explores how open models including NVIDIA's NEMO family of models serve as the primary catalyst for transparency, reproducibility, and democratisation in AI development.

By providing unfettered access to model parameters and training methodologies, open models allow researchers to conduct deep-dive audits into safety, bias, and mechanistic interpretability that are often impossible behind closed APIs. We will discuss how an open ecosystem prevents "innovation silos," fosters global collaboration, and ensures that the future of AI is shaped by a diverse academic community rather than a handful of corporate entities. The session concludes with a look at the emerging landscape of high-performance open models and their role in preserving the university's mission of public-interest research.