NVIDIA Nemotron Model Reasoning Challenge
Reasoning ability is increasingly important as AI systems take on more complex tasks across research and real-world applications. Models now need to solve harder problems, do it efficiently, and be reproducible so others can build on the work.
NVIDIA Nemotron is designed for exactly this: a family of open source models, datasets, and training recipes built to support efficient high-quality reasoning. Nemotron already performs competitively on leading reasoning benchmarks, and we want to see how far the community can push it.
This challenge will invite you to use Nemotron’s open resources - models, datasets, and techniques - to improve its reasoning benchmark performance. You’ll be free to explore your own approach: prompting, data filtering, synthetic data generation, tuning, etc.
Because Nemotron and its resources are openly available, your contributions don’t just improve a single benchmark run - they help strengthen an open ecosystem that others can study, reproduce, and extend. Making Nemotron better means helping the broader open source AI community move forward together.
More details on the competition format, timeline, and prizes are coming soon.
Register to this event for updates and to be notified when the challenge goes live.
In the meantime, start exploring...
Here are some resources to get familiar with Nemotron’s models, datasets, and workflows:
Nemotron Overview with links to explore models, datasets, developer tools, and starter kits
Join the Conversation
Bring your questions - we’d love to hear how you’re thinking about Nemotron and reasoning ahead of the challenge.
Visit the Nemotron Discord Channel and join our weekly "Nemotron Labs" livestreams, where NVIDIA experts walk through tools, demos, and research topics.