Nemotron 3

weights

open

data

open(~40M post-training/SFT+RL samples released as Nemotron-SFT-Data/Nemotron-RL-Data

code

open(training recipes + RL env configs)

license

NVIDIA-Nemotron-Open-Model-License(permissive,commercial,non-OSI)

NVIDIA releases the weights, post-training data and training recipes. Nano and Super ship under its own non-OSI Nemotron Open Model License and Ultra under the non-OSI OpenMDW License 1.1, and both permit commercial use. The pretraining side is thinner than NVIDIA's 'fully open' framing suggests: the specialized corpus is real data, but the code corpus ships only metadata describing it, and Ultra's code, math and multilingual data require gated approval.

• https://developer.nvidia.com/blog/introducing-nemotron-3-super-an-open-hybrid-mamba-transformer-moe-for-agentic-reasoning/ recorded 2026-08-13

  NVIDIA primary blog - "Nemotron 3 Super is fully open-weights, datasets, and recipes"; license = NVIDIA Nemotron Open Model License (non-OSI)

• https://huggingface.co/api/datasets/nvidia/Nemotron-Pretraining-Code-v3 recorded 2026-08-13

  67 files, 1.20GB, ungated, all under Nemotron-Code-Metadata/*.parquet -- metadata describing the code corpus, not the corpus itself

• https://huggingface.co/api/datasets/nvidia/Nemotron-Pretraining-Specialized-v1.2 recorded 2026-08-13

  92 files, 53.62GB, ungated. Nemotron-Pretraining-Fact-Seeking/*.parquet and related synthetic pretraining sets.

• https://research.nvidia.com/labs/nemotron/Nemotron-3/ recorded 2026-08-13

  Nemotron 3 Nano/Super/Ultra family; the Data section still lists Nemotron-SFT-Data ("Collection of new Nemotron 3 Nano SFT datasets") and Nemotron-RL-Data alongside the pretraining corpora, followed by a Model Recipes section

Adoption sums Hugging Face downloads across the three ungated BF16 repositories in the Nemotron 3 family: Nano, Super and Ultra. Quantized NVFP4, FP8 and GGUF builds are excluded to avoid double-counting the same weights, and the family is also distributed free through OpenRouter and NVIDIA's build/NIM platform, neither of which this count captures, so the reading is a floor rather than a ceiling.

• https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 recorded 2026-08-13

  914,120 downloads in the trailing 30 days, ungated

• https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16 recorded 2026-08-13

  870,782 downloads in the trailing 30 days, ungated

• https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 recorded 2026-08-13

  441,153 downloads in the trailing 30 days, ungated

• https://research.nvidia.com/labs/nemotron/Nemotron-3/ recorded 2026-08-13

  official distribution via NVIDIA build/NIM + HF for agentic use

Nemotron 3 Ultra, the family's largest model, sits in the same frontier tier as recent open models such as GLM-5.1, Kimi-K2.6 and DeepSeek-V4 in NVIDIA's own comparison, ahead on some benchmarks and behind on others; it trails them on SWE-Bench Verified (70.7 against 73.5 to 76.2). Super and Nano perform below it.

• https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Super-Technical-Report.pdf recorded 2026-08-13

  Agentic Reasoning Benchmark table, N-3-Super column - MMLU-Pro 83.73, AIME25 (no tools) 90.21, GPQA (no tools) 79.23, SWE-Bench (OpenHands) 60.47; the Qwen3.5-122B-A10B column on the same rows reads 86.70 and 66.40, and the GPT-OSS-120B column 81.00 and 41.9

• https://research.nvidia.com/labs/nemotron/Nemotron-3/ recorded 2026-08-13

  Nemotron 3 reasoning/agentic positioning - Nano "more accurate than GPT-OSS-20B and Qwen3-30B-A3B-Thinking-2507 on popular benchmarks"; Ultra released