Nvidia has significantly expanded its open model program in 2025, releasing more models and datasets, which positions them among a select few companies doing so at scale. In an interview with Bryan Catanzaro, VP of Applied Deep Learning Research at Nvidia, he highlighted the recent launch of the Nemotron 3 Nano model in December 2025, along with plans for larger Super and Ultra variants. Catanzaro explained that Nvidia’s strategy not only enhances their market presence but also supports developers and researchers who are pushing the boundaries of AI. He emphasized the importance of open AI models in fostering innovation across different sectors, drawing parallels to how the internet transformed various industries. Nvidia's approach includes a commitment to releasing high-quality datasets, which come with legal risks but are essential for training effective models. In 2025, Nvidia began offering an array of pretraining datasets, making them one of the few organizations to do so at this scale. Key releases include the CC-v2 collections and specialized datasets for various applications. The timeline of Nvidia's Nemotron models showcases their rapid development, from foundational work in 2019 to the latest innovations in 2025, reflecting a clear trajectory in advancing AI capabilities. Catanzaro's insights into the team culture at Nvidia reveal a focus on collaboration and creativity, which is integral to their success in building and releasing these open models.