Build the Future of AI in the Open
Open source advances through collaboration. NVIDIA builds frontier open models and contributes data, software, and standards across the AI ecosystem.
Introduction
Empowering Innovation
Open source is fundamental to making accelerated computing broadly accessible and usable across industries. NVIDIA champions and invests in open models and platforms to spark innovation, grow ecosystems, and empower enterprises to integrate and deploy AI with control and scale.
Ecosystem
Open Source Organizations NVIDIA Supports
Open source foundations are critical to ensuring the long-term sustainability, security, and governance of open source projects. NVIDIA supports many of these foundations and consortiums, both financially and through participation in governing boards and technical advisory boards or committees.
-
Funding
-
Upstream
-
Governance
NVIDIA Models
Open Models Built for the Frontier
Intelligence is more than just language. NVIDIA open model families span how machines reason, see the physical world, move through it, drive, and read biology.
Open Weights, Data, Recipes, and Evals
Being able to download a model doesn't mean much if you can't reproduce it, otherwise, it's just a licensed black box. NVIDIA publishes the full pipeline: the weights, the data, the recipe, and the tools to evaluate how to get there. Verify the results yourself—no need to take anyone's word for it.
| Model Family | Weights | Data | Recipe | Evals | License |
| Nemotron 3 | âś“ | âś“ | âś“ | âś“ | OpenMDW 1.1 |
| Cosmos 3 | âś“ | âś“ | âś“ | âś“ | OpenMDW 1.1 |
| Isaac GR00T | âś“ | âś“ | âś“ | âś“ | NVIDIA Open Model License, Apache 2.0 |
| Alpamayo | âś“ | âś“ | âś“ | âś“ | OpenMDW 1.1 |
| BioNeMo | âś“ | âś“ | âś“ | âś“ | NVIDIA Open Model License, Apache 2.0 |
| Earth-2 | âś“ | âś“ | âś“ | âś“ | NVIDIA Open Model License, Apache 2.0 |
| Ising | âś“ | âś“ | âś“ | âś“ | OpenMDW 1.1 |
Commitments
Committed to Collaboration
Projects
An Open Stack That Makes Open Models Usable
A model is only as open as your ability to run it. NVIDIA open sources the underlying tooling—training, serving, data curation, evaluation, and the drivers themselves.
NeMo
Develop, train, fine-tune, and evaluate generative AI models.
Megatron Core
Train large transformer models across distributed GPU systems.
RAPIDS
Accelerate data preparation, analytics, and machine learning.
Isaac Platform
An open robotics development platform for simulating, training and deploying robotics applications.
TensorRT-LLM
Optimize LLM inference on NVIDIA GPUs.
Dynamo
Serve generative AI models across distributed GPU infrastructure.
Triton Inference Server
Serve models from multiple AI frameworks at scale.
NCCL
Enable high-performance communication across GPUs and nodes.
CUTLASS
Build high-performance CUDA kernels with open C++ templates.
Open GPU Kernel Modules
Support open source Linux GPU driver development.
Slurm
A highly configurable open source workload and resource manager.
Resources
Join the NVIDIA Developer Community
News
Latest Open Source News
Next Steps
Ready to Get Started?
Engage with NVIDIA open projects, repositories, and developers.
Explore More Projects
Learn more about the open source projects NVIDIANs have worked on.