Nemotron 3 Super 120B-A12B has 123.6B parameters (12B active per token). In BF16 its weights alone take 225 GiB; quantized to 4-bit, about 75.5 GiB. Each 32K-token request adds 0.25 GiB of KV cache. Cheapest way to run it today: 4 × GeForce RTX 3090 (24 GB) at $0.572 per hour.
Read every day from Hugging Face. Split files are added together; vision projectors are left out.
How many GPUs to run Nemotron 3 Super 120B-A12B
Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.
GPU
BF16 / FP16
FP8 / INT8
4-bit (GGUF Q4_K_M)
GeForce RTX 4090 24 GB
16 (several servers)
6
4
GeForce RTX 5090 32 GB
8
4
3
RTX PRO 6000 Blackwell 96 GB
3
2
1
Mac M5 Max 128 GB 96 GB usable
3
2
1
NVIDIA H100 80 GB
4
2
2
NVIDIA H200 141 GB
2
1
1
NVIDIA B200 (HGX) 180 GB
2
1
1
Cheapest way to run Nemotron 3 Super 120B-A12B today
For each GPU: the fewest cards that fit Nemotron 3 Super 120B-A12B with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-10-10).
On-demand prices from RunPod, Vast.ai, Verda and Azure, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.
KV cache: how memory grows with context
Architecture: KV on 8 of 88 layers (GQA, 2 KV × 128) · 40 Mamba-2 layers · 40 MoE layers. Its KV cache is 11.0× smaller than if every layer used full attention, because only 8 of its 88 layers keep a KV cache; the others hold a small fixed-size state. Only 12B parameters are active per token, but all 124B must sit in memory: VRAM depends on the total, speed on the active part.
Context per request
BF16 cache
FP8 cache
8K tokens
0.06 GiB
0.03 GiB
32K tokens
0.25 GiB
0.13 GiB
128K tokens
1.00 GiB
0.50 GiB
256K tokens
2.00 GiB
1.00 GiB
Plus 162 MiB of recurrent state per request, whatever the context.
Nemotron 3 Super 120B-A12B VRAM FAQ
How much VRAM does Nemotron 3 Super 120B-A12B need?
In BF16 the weights alone take 225 GiB (241 GB). In FP8 that is 113 GiB, and about 75.5 GiB with 4-bit quantization (Q4_K_M). Each request then adds KV cache: 0.06 GiB at 8K tokens and 0.25 GiB at 32K (BF16 cache).
Can Nemotron 3 Super 120B-A12B run on a single RTX 4090 (24 GB)?
No. Even in 4-bit it needs 4 RTX 4090s. In 4-bit it fits on a single RTX PRO 6000 Blackwell (96 GB).
What is the cheapest way to run Nemotron 3 Super 120B-A12B?
On 2026-10-10, the cheapest on-demand setup is 4 × GeForce RTX 3090 (24 GB) with 4-bit (GGUF Q4_K_M) weights on Vast.ai, at $0.572 per hour (about $417.56 per month). Next: 2 × RTX 6000 Ada (48 GB) with 4-bit (GGUF Q4_K_M) weights on Vast.ai, at $0.988 per hour. Sized for one 8K-token request; prices are checked every hour.
How many H100 GPUs does Nemotron 3 Super 120B-A12B need?
With one request and an 8K-token context: 4 in BF16, 2 in FP8 and 2 in 4-bit (vLLM, 90% of memory usable). Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.
Nemotron 3 Super 120B-A12B VRAM badge
For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/nemotron-3-super.json.
[](https://studiotvai.com/vram-requirements/nemotron-3-super)
Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.
NVIDIA, United States · Released March 10, 2026