howtospark
models / nemotron-3-nano-30b-a3b

Nemotron 3 Nano 30B-A3B

nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
params
31.6B
active params
3.0B
sparsity
MoE
attention
SSM / hybrid
architecture
nemotron_h
license
other
downloads
1.1M
likes
791
hf updated
Mar 15, 2026

Visualize

total size

each column is a layer — lit blocks are the experts that fire per token

bits / param
16-bit
footprint
58.8 GiB
sparks to fit
1×
block ≈
247M params
active paramsdormant expertsghost = precision traded awayblock fill ∝ bits kept · Nemotron 3 Nano 30B-A3B ≈ 31.577937344B params

Eval scores (compare all)

Reported

78.3MMLU-Pro
73GPQA Diamond
68.3LiveCodeBench
33.3SciCode
10.6HLE
50MiniF2F