Nemotron 3 Nano 30B-A3B has 31.6B parameters (3B active per token). In BF16 its weights alone take 58.8 GiB; quantized to 4-bit, about 22.0 GiB. Each 32K-token request adds 0.19 GiB of KV cache. Cheapest way to run it today: 2 × GeForce RTX 3090 (24 GB) at $0.286 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 Nano 30B-A3B
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
3
2
2
GeForce RTX 5090 32 GB
3
2
1
RTX PRO 6000 Blackwell 96 GB
1
1
1
Mac M5 Max 128 GB 96 GB usable
1
1
1
NVIDIA H100 80 GB
1
1
1
NVIDIA H200 141 GB
1
1
1
NVIDIA B200 (HGX) 180 GB
1
1
1
Cheapest way to run Nemotron 3 Nano 30B-A3B today
For each GPU: the fewest cards that fit Nemotron 3 Nano 30B-A3B 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 6 of 52 layers (GQA, 2 KV × 128) · 23 Mamba-2 layers · 23 MoE layers. Its KV cache is 8.7× smaller than if every layer used full attention, because only 6 of its 52 layers keep a KV cache; the others hold a small fixed-size state. Only 3B parameters are active per token, but all 32B must sit in memory: VRAM depends on the total, speed on the active part.
Context per request
BF16 cache
FP8 cache
8K tokens
0.05 GiB
0.02 GiB
32K tokens
0.19 GiB
0.09 GiB
128K tokens
0.75 GiB
0.38 GiB
256K tokens
1.50 GiB
0.75 GiB
Plus 46.8 MiB of recurrent state per request, whatever the context.
Nemotron 3 Nano 30B-A3B VRAM FAQ
How much VRAM does Nemotron 3 Nano 30B-A3B need?
In BF16 the weights alone take 58.8 GiB (63 GB). In FP8 that is 30.1 GiB, and about 22.0 GiB with 4-bit quantization (Q4_K_M). Each request then adds KV cache: 0.05 GiB at 8K tokens and 0.19 GiB at 32K (BF16 cache).
Can Nemotron 3 Nano 30B-A3B run on a single RTX 4090 (24 GB)?
No. Even in 4-bit it needs 2 RTX 4090s. In 4-bit it fits on a single GeForce RTX 5090 (32 GB).
What is the cheapest way to run Nemotron 3 Nano 30B-A3B?
On 2026-10-10, the cheapest on-demand setup is 2 × GeForce RTX 3090 (24 GB) with FP8 / INT8 weights on Vast.ai, at $0.286 per hour (about $208.78 per month). Next: 1 × NVIDIA A100 40GB with FP8 / INT8 weights on Vast.ai, at $0.401 per hour. Sized for one 8K-token request; prices are checked every hour.
How many H100 GPUs does Nemotron 3 Nano 30B-A3B need?
With one request and an 8K-token context: 1 in BF16, 1 in FP8 and 1 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 Nano 30B-A3B 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-nano.json.
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 December 4, 2025