CanItRun Logocanitrun.

Nemotron 3 Super 120B

Nemotron 3 Super 120B needs roughly 68.0GB VRAM at Q4_K_M quantization (269.6GB at FP16). 28 GPUs we track can run it fully in VRAM at 8k context.

NVIDIA120B params12B active (MoE)1024k contextNVIDIACommercial use ok

VRAM at each quantization

Assumes 8k context. KV cache grows linearly with context length.

QuantWeightsKV cacheTotal
FP16240.0 GB0.74 GB269.6 GB
Q8120.0 GB0.74 GB135.2 GB
Q6_K90.0 GB0.74 GB101.6 GB
Q5_K_M75.0 GB0.74 GB84.8 GB
Q4_K_M60.0 GB0.74 GB68.0 GB
Q3_K_M48.0 GB0.74 GB54.6 GB
Q2_K36.0 GB0.74 GB41.1 GB

Benchmarks

Benchmarks for this model are not yet available on the Open LLM Leaderboard v2. This is common for recently released models. Check back soon.

GPUs that run Nemotron 3 Super 120B natively (28)

Plus 11 GPUs that run it with CPU offload (slower)

Notes

Hybrid Mamba-Transformer LatentMoE with 512 routed experts, 22 active per token. Trained on 25T tokens.

Hugging Face ↗Released 2026-03-11