NVIDIA
NVIDIA, headquartered in Santa Clara, develops the open-weight Nemotron family of AI models.
| Model | Released | Weights | Hosts | Cheapest host | Input | Output |
|---|---|---|---|---|---|---|
| Nemotron 3.5 Super VLfree now | Oct 2026 | Open · OpenMDW-1.1 | 7 | Crusoe | $0.10 | $0.40 |
| Nemotron 3.5 Lightningfree now−33% | Aug 2026 | Open · NVIDIA Open Model License | 4 | Io Net | $0.0469 | $0.13 |
| Nemotron 3 Ultrafree now | Jun 2026 | Open · NVIDIA Open Model License | 3 | DeepInfra | $0.50 | $2.20 |
| Nemotron 3.5 Content Safetyfree now | Jun 2026 | Open · NVIDIA Open Model License | 1 | DeepInfra | $0.20 | $0.20 |
| Nemotron 3 Nano Omni 30B A3B Reasoningfree now | Apr 2026 | Open · NVIDIA Open Model Agreement | — | no published price | — | — |
| Nemotron 3 Superfree now | Mar 2026 | Open · NVIDIA Open Model License | 4 | DeepInfra | $0.085 | $0.40 |
| Nemotron 3 Nano 30B A3B | Dec 2025 | Open · NVIDIA Open Model License | 4 | Crusoe | $0.05 | $0.20 |
7 of 7 models · prices per million tokens at the cheapest host, read Oct 23, 09:51 UTC
Deals now
- Free modelNemotron 3.5 Super VLat OpenRouter · until no end date
- Free modelNemotron 3.5 Lightningat OpenRouter · until no end date
- Free modelNemotron 3.5 Content Safetyat OpenRouter · until no end date
- Free modelNemotron 3 Ultraat OpenRouter · until no end date
- Free modelNemotron 3 Nano Omniat OpenRouter · until no end date
- Free modelNemotron 3 Superat OpenRouter · until no end date
- −33%Nemotron 3.5 Lightningat Io Net
About NVIDIA
NVIDIA was founded on 5 April 1993 by Jensen Huang, Chris Malachowsky and Curtis Priem.1, 7, 8, 9
Huang previously worked at LSI Logic and AMD, and Malachowsky held engineering roles at HP and Sun Microsystems.1, 7, 8, 9
The company, incorporated in California and headquartered in Santa Clara, is led by Huang as president and chief executive officer.1, 2, 7, 8
NVIDIA published its Nemotron-3 8B family of foundation models on 15 November 2023, aimed at enterprises building custom language models.5, 6
On 14 June 2024 it announced Nemotron-4 340B, a family of open models for generating synthetic data to train language models.11
The Nemotron 3 family followed on 15 December 2025, with Nemotron 3 Super, which has 12 billion active and 120 billion total parameters, on 10 March 2026.12
Nemotron 3 Ultra, with 55 billion active and 550 billion total parameters, was released on 4 June 2026.12
NVIDIA then published Nemotron 3.5 Super and Lightning on 14 October 2026 and Nemotron 3.5 Bolt on 15 October 2026.12
Its research site says the project aims to help NVIDIA build and deploy future AI systems and to contribute to the openly developed AI community.13
NVIDIA describes the family as open models with open weights, under licences that vary by model, including OpenMDW 1.1 for Nemotron 3 Ultra.3, 4, 5
NVIDIA says it is located in California and in most cases stores user information in the United States, and its developer terms are governed by Delaware and US law.2, 10
Official links
- Website nvidia.com
- API docs docs.api.nvidia.com
- GitHub github.com/NVIDIA
- GitHub github.com/NVIDIA-NeMo
- Hugging Face huggingface.co/nvidia
- Blog blogs.nvidia.com
- X twitter.com/nvidia
- X x.com/NVIDIAAI
- LinkedIn linkedin.com/company/nvidia
Sources
- NVIDIA Corporation Form 10-K for the fiscal year ended January 25, 2026 · sec.gov · read Oct 23
- Terms of Use | NVIDIA Developer · developer.nvidia.com · read Oct 23
- Nemotron AI Models | NVIDIA Developer · developer.nvidia.com · read Oct 23
- nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 · Hugging Face · huggingface.co · read Oct 23
- nvidia/nemotron-3-8b-base-4k · Hugging Face · huggingface.co · read Oct 23
- NVIDIA AI Foundation Models: Build Custom Enterprise Chatbots and Co-Pilots with Production-Ready LLMs | NVIDIA Technical Blog · developer.nvidia.com · read Oct 23
- Our History: Innovations Over the Years | NVIDIA · nvidia.com · read Oct 23
- Jensen Huang | NVIDIA Newsroom · nvidianews.nvidia.com · read Oct 23
- Chris A. Malachowsky | NVIDIA Newsroom · nvidianews.nvidia.com · read Oct 23
- NVIDIA Privacy Policy · nvidia.com · read Oct 23
- NVIDIA Releases Open Synthetic Data Generation Pipeline for Training Large Language Models | NVIDIA Blog · blogs.nvidia.com · read Oct 23
- Projects - NVIDIA Nemotron · research.nvidia.com · read Oct 23
- NVIDIA Nemotron Research - NVIDIA Nemotron · research.nvidia.com · read Oct 23
Every fact above links to its source and the date it was read; last checked Oct 23. Not affiliated with NVIDIA. Corrections or removal on request: contact.