NVIDIA Nemotron
Benchmark-topping open reasoning models, built with transparent training data, that think fast and lower inference cost.
Overview
What Is NVIDIA Nemotron?
The NVIDIA Nemotron™ family of multimodal models provides state-of-the-art reasoning models specifically designed for enterprise-ready AI agents. These models excel in graduate-level scientific reasoning, advanced math, coding, instruction following, tool calling, and visual reasoning.
Nemotron models are trained with transparent, open-sourced training data—giving enterprises full visibility, enabling better compliance, and ensuring trustworthy AI deployment.
They are optimized for a range of computing platforms: Nano for cost-efficiency and edge deployment, Super for balanced accuracy and compute efficiency on a single GPU, and Ultra for maximum accuracy in data centers.
The Nemotron models are commercially viable with an open license that allows for customization and data control, and they can be deployed anywhere with NVIDIA NIM™ microservices.
Benefits
What Does Nemotron Bring to Agentic AI?
High Accuracy
Built on popular open reasoning models for their exceptional knowledge, post-trained with high-quality training data, and aligned to reason like humans, Nemotron models achieve the highest accuracy on leading benchmarks.
High Compute Efficiency
Through the pruning of larger models, the Nemotron family is optimized for top compute efficiency, using NVIDIA TensorRT™-LLM to deliver higher throughput and on-or-off reasoning capabilities.
Commercially Viable
NVIDIA’s post-training data and optimization techniques ensure powerful, transparent, and adaptable models for developers and enterprises. Models and training data are published openly on Hugging Face.
Secure and Simple Deployment
The Nemotron model family, available as optimized NIM microservices, offers peak inference performance and flexible deployment options, ensuring superior security, privacy, and portability.
Models
Models for Diverse Workloads
Nemotron models excel in vision for enterprise optical character recognition (OCR) and in reasoning for building agentic AI. Research models are also available for experimentation and customization.
Technology
Building Blocks for Agentic AI
Start building AI agents with NVIDIA NeMo™ for custom agentic AI, NVIDIA NIM for fast, enterprise-ready deployment, and NVIDIA Blueprints for accelerating development with customizable reference workflows.
Starting Options
Ways to Get Started With Nemotron
Adopters
Enterprises Using Nemotron
Resources
Explore the Latest in Nemotron
Blogs
FAQs
NVIDIA Nemotron models aren't just open, but truly open source. NVIDIA publishes the training datasets, techniques, and model weights so the open-source community can benefit from our learnings and use these resources to create their own models.
The NVIDIA Open Model License is a permissive license that allows users to use, modify, distribute, and commercially deploy the models and derivatives without crediting NVIDIA, to encourage innovation and further development of generative AI.
Yes, you can download and run NVIDIA Nemotron models from Hugging Face for free in production.
NVIDIA also offers Nemotron models as NVIDIA NIM microservices for secure, scalable deployment, which requires an NVIDIA AI Enterprise license. You can try the Nemotron models and download the NIM microservices from build.nvidia.com.
Yes, NVIDIA is committed to publishing more Nemotron models, datasets, and techniques to enable open-source ecosystems.
NVIDIA Nemotron models are built on top of frontier open models, making it possible to build better models faster. Additionally, NVIDIA publishes the model weights, training datasets, and training techniques so the developer community can use these different parts of Nemotron to train their own models.
Yes. NVIDIA built the Llama Nemotron models on top of the Llama model family using NVIDIA’s open datasets and advanced techniques, such as Neural Architecture Search (NAS). The Llama Nemotron models inherit the parent Llama model license.
NVIDIA provides a variety of tools, such as NVIDIA Dynamo, TensorRT-LLM, and NIM, to run Nemotron models at scale in production. You can also use popular open-source libraries, such as SGLang and vLLM.
Next Steps
Ready to Get Started?
Use the right tools and technologies to take NVIDIA Nemotron models from development to production.
Get in Touch
Talk to an NVIDIA product specialist about moving from pilot to production with the security, API stability, and support that comes with NVIDIA AI Enterprise.
Stay Up to Date on NVIDIA Agentic AI News
Get the latest agentic AI news, technologies, breakthroughs, and more sent straight to your inbox.