Nvidia has unveiled the Nemotron Coalition, partnering with eight AI research groups to jointly develop open frontier AI models on its DGX Cloud platform. This collaboration aims to fuel the upcoming Nemotron 4 family of AI models, combining expertise from companies including Black Forest Labs, Cursor, LangChain, Mistral AI, Perplexity, Reflection AI, Sarvam, and Thinking Machines Lab – the latter led by Mira Murati, ex-CTO of OpenAI.
The coalition’s initial project is a base model co-created by Nvidia and Mistral AI, trained on the DGX Cloud infrastructure. The other members will contribute additional datasets, evaluation metrics, and domain specialization during the model’s post-training refinement. Upon completion, Nvidia plans to open-source this foundational model, which will serve as the core for the broader Nemotron 4 lineup, potentially powering a spectrum of AI applications.
Alongside these announcements, Nvidia revealed a suite of novel AI models spanning sectors such as agentic AI, robotics, autonomous vehicles, and drug development. Among them is Cosmos 3, a foundational AI model designed to integrate synthetic environment creation with physical reasoning capabilities, scheduled for release later this year.
In the biotechnology space, the Proteina-Complexa model, part of Nvidia’s BioNeMo platform, targets the design of protein binders essential to drug discovery. Early adopters include pharmaceutical and biotech companies like Novo Nordisk, Viva Biotech, and Manifold Bio, highlighting Nvidia’s expansion into life sciences AI applications.
Nvidia’s strategy to foster an ecosystem of open frontier AI models comes amid intense competition with tech giants and startups alike, all pushing aggressively into foundational AI technologies. By pooling knowledge and resources from varied AI specialists, Nvidia hopes to accelerate innovation in areas that require highly specialized AI models-offering a challenge to both open-source and proprietary AI efforts.
The involvement of Mira Murati’s Thinking Machines Lab highlights the coalition’s ambition. After her high-profile exit from OpenAI, her new venture’s inclusion signals a shift toward broader collaborative AI model development, moving beyond the siloed approaches of earlier years.
Nemotron 4 and its supporting models will likely be scrutinized not only for performance but also for how open they become after release, as industry observers look to see whether Nvidia can balance delivering advanced AI results with accessible AI infrastructure.