What GPU do I need to run nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8?
123.6B parameters, published in F8_E4M3. View on Hugging Face
Set up NVIDIA-Nemo…er-120B-A12B-FP8NVIDIA-Nemotron-3-Super-120B-A12B-FP8 is published by nvidia on Hugging Face, with 125,620 downloads and 278 likes to date. It's a NemotronHForCausalLM model built for text-generation, published natively in F8_E4M3.
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision, plus a fixed overhead for KV-cache, activations, and fragmentation. Full formula and assumptions: methodology.
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run NVIDIA-Nemotron-3-Super-120B-A12B-FP8 at its published (F8_E4M3) precision: 7× RTX 4000 SFF Ada on runpod, at $0.180/hr per GPU ($1.26/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
NVIDIA-Nemotron-3-Super-120B-A12B-FP8: common questions
Can NVIDIA-Nemotron-3-Super-120B-A12B-FP8 run on a single GPU?
No. At FP8 (native) it needs 138.1 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 20.0 GB RTX 4000 SFF Ada, and it takes 7 of them.
Is NVIDIA-Nemotron-3-Super-120B-A12B-FP8 already quantized?
Yes. It is published in FP8, one byte per parameter, so the 138.1 GB figure is already a quantized footprint rather than a full-precision one. Only INT4 goes below it, at 69.1 GB. A GPU without FP8 tensor cores cannot run it as published, which is why cards here are matched on precision support and not on VRAM alone.
How many GPUs do I need to run NVIDIA-Nemotron-3-Super-120B-A12B-FP8?
7 at FP8 (native). It needs 138.1 GB of VRAM and the cheapest capable live offer is a 20.0 GB RTX 4000 SFF Ada on runpod, so 7 of them come to $1.26/hr in total.
Does quantizing NVIDIA-Nemotron-3-Super-120B-A12B-FP8 lower the GPU bill?
Yes. At FP8 (native) the cheapest live fit is 7 RTX 4000 SFF Ada cards on runpod at $1.26/hr. At INT4 (quantized) it drops to one A100 on runpod at $1.19/hr, provided a quantized checkpoint exists for it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More nvidia models
- Gemma-4-31B-IT-NVFP4 (20.9B, BF16)
- parakeet-ctc-1.1b (1.1B, F32)
- NVIDIA-Nemotron-3-Super-120B-A12B-BF16 (123.6B, BF16)
- Cosmos-Reason2-2B (2.4B, BF16)
- nemotron-3.5-asr-streaming-0.6b (638M, F32)
- NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 (31.6B, BF16)
