SandRise
Ep 627 Research Paper 1:32 w/ Vince & Ava

How Open Models Are Driving AI Research

NVIDIA's open models, particularly Nemotron, Cosmos, and BioNeMo, are driving AI research by providing foundational tools for new studies, with 145 papers citing Nemotron at ICML 2026.

New ModelsNvidiaNemotronCosmosSynthetic Data Generation For ValidationPredictive ModelingResearch Paper
Read the source → Plain-text transcript →
Embed this episode

Paste this on any site — the player is a self-contained iframe with no cookies or trackers. 

<iframe src="https://sandrise.io/exploring-next/embed/627"
  width="100%" height="180" style="max-width:640px;border:0;border-radius:12px;overflow:hidden"
  title="Exploring Next — Episode 627 audio player"
  loading="lazy" allow="autoplay" referrerpolicy="strict-origin-when-cross-origin"></iframe>
Embed & API docs →
Script Llama 4 Scout Voice Hume Octave 2

Transcript

Ava So it seems like NVIDIA's open models are really making a splash in AI research. The article claims 145 papers at ICML 2026 are using their Nemotron models. That's a pretty big deal.

Vince Yeah, and it's not just Nemotron. They're also seeing a lot of use with their Cosmos and BioNeMo models. This is a clear indication that open models are becoming foundational to how AI research gets done.

Ava I'm a bit skeptical about the extent of their impact. I mean, 145 papers is a lot, but ICML accepts thousands of papers. What's the real influence here?

Vince Well, it's not just about the number of papers. It's about the areas they're being used in. Robotics, autonomous vehicles, biomedical research... these are all significant fields. And it's not just NVIDIA's research labs using them, but also other companies like Basecamp Research and Merck.

Ava That's a good point. I can see why this would be important for researchers who need reliable access to open inference. But what about the technical details? Are these models actually delivering on their promises?

Vince From what I can see, the models are being used for a range of applications, from predicting protein function to generating synthetic data. And there are some specific examples, like DreamDojo, which is using NVIDIA Cosmos to predict how robots would handle objects in new environments.

Ava Okay, I think I'm starting to see the picture. This isn't just about NVIDIA's models being popular, it's about them enabling actual research progress. I'd still like to see more technical details, but I can see why this matters.