49 models across 13 families · Latest: Cosmos 3 Nano (2026-05)
Accelerated AI for enterprise solutions
NVIDIA AI's portfolio covers 46 active models across 12 non-obsolete families, with task labels spanning coding, rag, and agents. Open a model detail page to compare provider routes and sourced benchmarks.
Portfolio context: 7 decision-task tags, 46 active tracked models, latest research stamp 2026-06-01.
Use this portfolio page for
- Teams evaluating coding, rag, and agents across this lab's releases
- Readers comparing families before locking a flagship SKU
- 46 tracked SKUs for migration and pricing follow-ups
Do not stop here for
- Choosing a hosting provider without opening a model page for price ladders
Active models
46
Non-deprecated SKUs linked to this researcher
Active families
12
Non-obsolete families in coverage
Open catalog
7 OSS
0 open-weight (text match)
Decision task tags
7
Mapped to the site-wide task taxonomy
Latest dated release
2026-05-31
Cosmos 3 Nano
Freshness
2026-06-01
Researched 3d ago
Release cadence
Showing 5 recent dated ships (full timeline below). Latest spotlight: Cosmos 3 Nano (2026-05-31).
Where this lab wins
- Coding: 1 tracked model with SWE-bench / HumanEval-style scores.
- RAG: 2 tracked models with ruler / needle retrieval benchmarks.
- Agentic: 2 tracked models with BFCL, tau-bench, and SWE-bench tool-use coverage.
- Long-context: 10 tracked models with context-token or InfiniteBench-class signal.
Flagship quality / price signal
Anchor SKU: Nemotron 3 Nano (best sourced coding Q/$ in this portfolio).
Quality / dollar unavailable for this anchor — missing benchmark coverage and/or output token price on the cheapest ladder route (open the model detail after pricing lands).
NVIDIA AI is an American AI research organization founded in 2015. Accelerated AI for enterprise solutions. NVIDIA AI ships 13 model families totaling 49 models, with the most recent release Cosmos 3 Nano in 2026-05. Notable families include Cosmos 3, Nemotron 3, and Nemotron-Cascade. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are. View official API endpoints, benchmark performance, and coding/agent fit for every NVIDIA AI model.
About
NVIDIA's journey into the realm of artificial intelligence, specifically in the areas of generative AI and large language models (LLMs), is marked by a series of strategic innovations that have equipped the company to lead in the AI landscape. Originally renowned for its high-quality graphics processing units (GPUs) used in gaming and multimedia, NVIDIA shifted gears to address the demands of AI, focusing on quality over sheer volume. This strategic pivot, underscored by the development of pioneering technologies, has attracted tech giants such as Microsoft, Amazon, and Facebook as customers, and has become instrumental in powering large-scale AI applications. NVIDIA's GPUs play a pivotal role in supporting the infrastructure required for complex AI tasks, as evidenced by Microsoft's significant expenditure on NVIDIA's chips 3. The company's engagement with AI began in earnest in the 2010s, laying the foundation for its success in generative AI. The launch of specialized GPUs, like the Tesla series and the introduction of significant architectures like Kepler, marked its early contributions. Additionally, the introduction of CUDA (Compute Unified Device Architecture) in 2006 was a landmark development that unlocked the parallel processing potential of GPUs, extending their applicability beyond traditional graphics tasks to include a wide array of AI applications 9. In 2014, NVIDIA further entrenched its position by introducing the cuDNN (CUDA Deep Neural Network) library, optimizing codes for deep learning models and significantly enhancing training and inference processes 9. Beyond hardware, NVIDIA's commitment to the broader AI ecosystem is evident through initiatives like the NVIDIA Deep Learning Institute (NDLI) and the integration of open-source frameworks. These efforts have cultivated a thriving developer community and accelerated the adoption of NVIDIA's technologies across diverse sectors. The development of the NeMo framework exemplifies this approach; NeMo offers a comprehensive platform for creating custom generative AI, including LLMs and vision language models (VLMs), streamlining the training and deployment of LLMs and making them accessible for global enterprises 1012. NVIDIA's impact on generative AI and LLMs is profound. Its technologies facilitate a plethora of applications from AI-driven video and image generation to the deployment of large language models and recommendation systems 48. Platforms like NVIDIA Omniverse showcase the company’s dedication to pushing AI’s boundaries, particularly in supporting diverse AI applications across industries. Ongoing advancements, such as those in the NeMo framework that offer significant speed enhancements for training LLMs, underscore NVIDIA's constant push for innovation. NVIDIA's remarkable success is a testament to its strategic foresight, technical expertise, and unwavering commitment to fostering a vibrant AI ecosystem.
Featured models
| Model | Released | Context | Input price ($/1M) | Output price ($/1M) | License |
|---|---|---|---|---|---|
| Cosmos 3 Nano | 2026-05-31 | 256k | - | - | OpenMDW 1.1 |
| Cosmos 3 Super | 2026-05-31 | 256k | - | - | OpenMDW 1.1 |
| Cosmos 3 Super Text2Image | 2026-05-31 | 4k | - | - | OpenMDW 1.1 |
Model families
Recent releases
- Cosmos 3 Nano- 2026-05-31
- Cosmos 3 Super- 2026-05-31
- Cosmos 3 Super Text2Image- 2026-05-31
- Cosmos 3 Super Image2Video- 2026-05-31
- Cosmos 3 Nano Policy DROID- 2026-05-31
FAQ
Who founded NVIDIA AI and when?
NVIDIA AI was founded in 2015 and is associated with Santa Clara, California, United States.
What models has NVIDIA AI released?
NVIDIA AI ships 49 models across 13 families: Cosmos 3, Nemotron 3, and Nemotron-Cascade.
Is NVIDIA AI's technology open source?
Some NVIDIA AI models are open-weight (Cosmos 3 Nano, Cosmos 3 Super, and Cosmos 3 Super Text2Image); others are proprietary (Nemotron 3 Nano 30B-A3B, Nemotron 3 Ultra, and NVIDIA Nemotron Nano 12B v2 VL BF16).
Where is NVIDIA AI headquartered?
NVIDIA AI is headquartered in Santa Clara, California, United States.
What is NVIDIA AI known for?
Accelerated AI for enterprise solutions. Its most prominent tracked family is Cosmos 3.
How can I access NVIDIA AI's models?
NVIDIA AI's models are available via AWS Bedrock, Cloudflare Workers AI, DeepInfra, Fireworks AI, and Microsoft Foundry.
Explore related pages
Last reviewed: 2026-06-01. Data sourced from public lab announcements and provider documentation.