Skip to content
AI Atlas
ArtifactquantizationBF16/F32/F8_E4M3/U8Open weightsof NVIDIA Nemotron 3.5 Lightning 30B A3B

nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4

published by NVIDIAhuggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-

This is a quantization of NVIDIA Nemotron 3.5 Lightning 30B A3B, not an independent model. Parameters, benchmarks, prices and lineage are recorded on the canonical model. Open NVIDIA Nemotron 3.5 Lightning 30B A3B →

data quality42

Updated 2 h ago · first seen 11 Sept 2026

model_01M294ZPF4P6M25JJ1HK86T3BD

File size
Format
Downloads
Published

Artifact facts

Weights dtype

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium

File size

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium

Downloads

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 6 h agomedium

Likes

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 6 h agomedium

Hugging Face repo

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

Pipeline tag

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

Library name

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium

Gated

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium

License

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium

Release date

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium

Last modified

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

Tags

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium

Hardware fit (this packaging)

Estimated

37 of 37 device × quantization combinations fit.

Run locally: your machine →
Estimated hardware fit
HardwareQuantizationDevice memoryEst. memoryFits
Apple M3 Ultra4bit—10.8 GB est.Yes
Mac Studio (Apple M5 Ultra)4bit—10.8 GB est.Yes
Apple M2 Ultra4bit—10.8 GB est.Yes
Apple M1 Ultra4bit—10.8 GB est.Yes
Apple M3 Max4bit—10.8 GB est.Yes
Apple M4 Max4bit—10.8 GB est.Yes
Mac Studio (Apple M5 Max)4bit—10.8 GB est.Yes
MacBook Pro (Apple M5 Max)4bit—10.8 GB est.Yes
Apple M2 Max4bit—10.8 GB est.Yes
Apple M1 Max4bit—10.8 GB est.Yes
Apple M4 Pro4bit—10.8 GB est.Yes
Mac mini (Apple M5 Pro)4bit—10.8 GB est.Yes
MacBook Pro (Apple M5 Pro)4bit—10.8 GB est.Yes
Apple M3 Pro4bit—10.8 GB est.Yes
Apple M1 Pro4bit—10.8 GB est.Yes
Apple M2 Pro4bit—10.8 GB est.Yes
Apple M44bit—10.8 GB est.Yes
iMac (Apple M4)4bit—10.8 GB est.Yes
Mac mini (Apple M6)4bit—10.8 GB est.Yes
MacBook Air (Apple M5)4bit—10.8 GB est.Yes
MacBook Pro (Apple M5)4bit—10.8 GB est.Yes
Apple M24bit—10.8 GB est.Yes
Apple M34bit—10.8 GB est.Yes
Apple M14bit—10.8 GB est.Yes
NVIDIA GeForce RTX 30904bit24 GB10.8 GB est.Yes
NVIDIA GeForce RTX 40904bit24 GB10.8 GB est.Yes
NVIDIA GeForce RTX 50904bit32 GB10.8 GB est.Yes
NVIDIA A100 80GB4bit80 GB10.8 GB est.Yes
NVIDIA H100 SXM4bit80 GB10.8 GB est.Yes
NVIDIA H100 NVL4bit94 GB10.8 GB est.Yes
NVIDIA DGX Spark4bit128 GB10.8 GB est.Yes
NVIDIA H2004bit141 GB10.8 GB est.Yes
NVIDIA H200 NVL4bit141 GB10.8 GB est.Yes
NVIDIA B2004bit180 GB10.8 GB est.Yes
AMD Instinct MI300X4bit192 GB10.8 GB est.Yes
AMD Instinct MI325X4bit256 GB10.8 GB est.Yes
NVIDIA DGX B2004bit1,440 GB10.8 GB est.Yes
Assumptions (7)
  • Estimated, not measured: weights = parameters × bytes/param × 1.15 runtime overhead (or the observed artifact file size when one is recorded).
  • bytes/param: 4bit = 0.5, 8bit = 1.0, fp16 = 2.0 (uniform quantization, no per-layer exceptions).
  • KV cache: 2 × layers × kv_heads × head_dim × 2 bytes × context × batch when the architecture is known; otherwise 0.5 GB per 8 192 tokens (× batch), independent of architecture (GQA/MLA models need less).
  • A model 'fits' when the estimate is at most the device memory minus 2 GB reserved for the OS and framework.
  • Mixture-of-experts models are estimated on total parameters (all experts must be resident); active parameters are ignored.
  • Device memory uses the largest configuration when several are listed (e.g. Apple silicon tiers).
  • Multi-GPU: device memories are summed; interconnect bandwidth, tensor-parallel replication and pipeline bubbles are not modelled.

Provenance

Attributed facts

24

Source tiers

T224

Freshest observation

2 h ago

Conflicts

None

Source documents
SourceDocumentTypeTierLast observedSnapshots
Hugging Face Hub (public pages, model cards, papers)huggingface.co/models?author=nvidia&p=0&sort=downloads listingT2· Quality secondary2 h ago7
Hugging Face Hub (public pages, model cards, papers)huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4/raw/main/README.md model_cardT2· Quality secondary5 h ago1
Hugging Face Hub (public pages, model cards, papers)huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 model_pageT2· Quality secondary6 h ago4

Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.