Uploaded April 2026 | Updated September 2026, 2 weeks ago
The NVIDIA Cosmos Cookoff brought together over 1,600 participants in a four-week virtual hackathon, sponsored by Nebius and Milestone Systems, to push the boundaries of physical AI.
Participants tackled a diverse range of challenges from vision AI applications to autonomous vehicle scenarios. They demonstrated remarkable depth by leveraging LLMs and fine-tuning techniques with the Cosmos Reason 2 model for advanced reasoning tasks.
This livestream showcases innovations from winning teams, highlighting both creativity and technical excellence. We also share the latest Cosmos updates unveiled at NVIDIA GTC, offering a glimpse into what’s next for the ecosystem.
What you'll learn:
- Practical approaches to using Cosmos Reason 2 with LLMs and fine-tuning for reasoning-heavy tasks
- What differentiated the winning projects—from idea to execution
- Emerging patterns and best practices from 1,600+ participants
- The latest Cosmos updates and roadmap highlights shared at NVIDIA GTC
🧑🏻🍳Read the Cosmos Cookbook → nvda.ws/4qevli8
📚 Explore Models & Datasets on GitHub → github.com/nvidia-cosmos
⬇️ Download Cosmos on Hugging Face → huggingface.co/collections/nvidia/nvidia-cosmos-2
👥 Join the Cosmos Community → discord.com/invite/nvidiaomniverse
🗳️ Contribute to the Cosmos Cookbook → nvda.ws/4aQcBkk
The NVIDIA Cosmos Cookoff brought together over 1,600 participants in a four-week virtual hackathon, sponsored by Nebius and Milestone Systems, to push the boundaries of physical AI.
Participants tackled a diverse range of challenges from vision AI applications to autonomous vehicle scenarios. They demonstrated remarkable depth by leveraging LLMs and fine-tuning techniques with the Cosmos Reason 2 model for advanced reasoning tasks.
This livestream showcases innovations from winning teams, highlighting both creativity and technical excellence. We also share the latest Cosmos updates unveiled at NVIDIA GTC, offering a glimpse into what’s next for the ecosystem.
What you'll learn:
- Practical approaches to using Cosmos Reason 2 with LLMs and fine-tuning for reasoning-heavy tasks
- What differentiated the winning projects—from idea to execution
- Emerging patterns and best practices from 1,600+ participants
- The latest Cosmos updates and roadmap highlights shared at NVIDIA GTC
🧑🏻🍳Read the Cosmos Cookbook → nvda.ws/4qevli8
📚 Explore Models & Datasets on GitHub → github.com/nvidia-cosmos
⬇️ Download Cosmos on Hugging Face → huggingface.co/collections/nvidia/nvidia-cosmos-2
👥 Join the Cosmos Community → discord.com/invite/nvidiaomniverse
🗳️ Contribute to the Cosmos Cookbook → nvda.ws/4aQcBkk
