Nemotron Nano 9B v2 · 33 GPUs measured first-party · llama.cpp Q4_K_M · Updated October 2026

What GPU Do You Need for Nemotron Nano 9B v2?

We ran Nemotron Nano 9B v2 (Q4_K_M, llama.cpp) on 33 GPUs, from budget cards to datacenter parts, and measured generation speed, prompt processing, power draw and peak VRAM on every one. The fastest was NVIDIA H100 80GB HBM3 at 227.7 tok/s.

Benchmarked weights: bartowski/nvidia_NVIDIA-Nemotron-Nano-9B-v2-GGUF

Fastest we measured

NVIDIA H100 80GB HBM3

227.7 tok/s on Nemotron Nano 9B v2, the ceiling. Measured on our bench. 80GB of VRAM, $30,000 at launch.

Pros

• 227.7 tok/s on Nemotron Nano 9B v2
• 80GB, clears the Nemotron Nano 9B v2 floor
• Rentable by the hour rather than bought

Cons

• 700W board rating
• Datacenter or workstation hardware, not a retail purchase

Best consumer card

NVIDIA GeForce RTX 5090

202.7 tok/s on Nemotron Nano 9B v2, fastest card you can buy at retail. Measured on our bench. 32GB of VRAM, $1,999 at launch.

Pros

• 202.7 tok/s on Nemotron Nano 9B v2
• 32GB, clears the Nemotron Nano 9B v2 floor

Cons

• 575W board rating

Cheapest card that runs it

GeForce RTX 4060

37.02 tok/s on Nemotron Nano 9B v2, lowest launch price that still fits. Measured on our bench. 8GB of VRAM, $299 at launch.

Pros

• 37.02 tok/s on Nemotron Nano 9B v2
• 8GB, clears the Nemotron Nano 9B v2 floor

Cons

• 115W board rating

Best value

GeForce RTX 5070

86.81 tok/s on Nemotron Nano 9B v2, most speed per dollar. Measured on our bench. 12GB of VRAM, $549 at launch. That is 158.1 tok/s per $1,000 of launch price.

Pros

• 86.81 tok/s on Nemotron Nano 9B v2
• 12GB, clears the Nemotron Nano 9B v2 floor

Cons

• 250W board rating

What the numbers show. Across 33 GPUs measured on our own bench, H100 80GB HBM3 is fastest at 228 tok/s. Per dollar of launch price, RTX 5060 Ti gives the most (179.3 tok/s per $1,000). The fastest card with 16GB or less is RTX 5080 at 120 tok/s. The measured peak was ~7GB, so it runs on 8GB cards and up. That floor is a property of the model, so it applies to every GPU, measured or not.

About Nemotron Nano 9B v2. Nemotron Nano 9B v2: from nvidia, 8.9B parameters, on Hugging Face since August 2025. 538,129 downloads in the last 30 days and 1 community quantizations.

How it compares. H100 80GB HBM3: Nemotron Nano 9B v2 227.7 tok/s, gemma-2-9b 174.7 (9B), Qwen2.5-VL 7B Instruct 267.7 (8B), Qwen3 8B 244.2 (8B), DeepSeek-R1-0528-Qwen3-8B 245.3 (8B). 3 of 4 beat Nemotron Nano 9B v2 here.

Cost on a rented GPU. 1M generated tokens of Nemotron Nano 9B v2: $0.25 on a RTX 3080 ($0.082/hr, 3.0 hours), $2.61 on a H100 80GB HBM3 ($2.14/hr, 73 min, 10.4x the cost).

Speed tiers for Nemotron Nano 9B v2. 30+ tok/s: 33 (RTX 5090, RTX 4090, RTX 5080). 30 tok/s is roughly where replies outpace reading.

Reading your prompt. Before Nemotron Nano 9B v2 writes anything it reads the input: 11166.2 tok/s on the RTX 5090 (0.4s for a 4,000-token prompt), 7951.1 on the RTX 4090 (0.5s), 932.2 on the T4 (4.3s). Long documents and big code files feel this number more than the generation speed.

VRAM for Nemotron Nano 9B v2. Measured peak 6.3GB, so 8GB is the smallest common card size; smallest card it ran on: RTX 3070 Ti (8GB). With long context: Q4_K_M 8GB (tested), Q2_K 7GB, Q3_K_M 7GB, Q5_K_M 9GB, Q6_K 12GB.

Power on Nemotron Nano 9B v2. Most efficient: H100 80GB HBM3, 268W, 0.33 kWh per 1M generated tokens. Hungriest: RTX 3090 Ti, 303W, 0.70 kWh. At $0.15/kWh: $0.049 per 1M generated tokens.

Our verdict

Fastest on Nemotron Nano 9B v2: NVIDIA H100 80GB HBM3, 227.7 tok/s. Best desktop card: NVIDIA GeForce RTX 5090, 202.7 tok/s. Cheapest consumer card that ran it: GeForce RTX 4060 ($299, 37.02 tok/s). Cheapest to rent per job: NVIDIA GeForce RTX 3080, $0.25 per 1M generated tokens.

FAQ

What GPU do I need to run Nemotron Nano 9B v2?

About 7GB. Cheapest consumer card that ran it: GeForce RTX 4060 (8GB, 37.02 tok/s).

How fast is Nemotron Nano 9B v2 on the NVIDIA GeForce RTX 5090?

202.7 tok/s at 291W, 89% of the NVIDIA H100 80GB HBM3.

How much does it cost to run Nemotron Nano 9B v2 in the cloud?

$0.25 per 1M generated tokens on a NVIDIA GeForce RTX 3080 at $0.082/hr, cheapest of 28 rentable cards we measured.

Can I run Nemotron Nano 9B v2 on a 12GB, 16GB or 24GB card?

It used 6.3GB at the precision we tested. 12GB: yes; 16GB: yes; 24GB: yes.

Is the RTX 4090 or the RTX 3090 faster for Nemotron Nano 9B v2?

The RTX 4090: 121.9 vs 108.9 tok/s, 12% faster on our bench.

Should I buy or rent a RTX 5090 for Nemotron Nano 9B v2?

Its $1,999 launch price buys 5,139 rented hours at $0.39/hr, enough for about 3,750x 1M generated tokens of Nemotron Nano 9B v2. Buy only if you'll run more than that.

How we test

llama.cpp llama-bench at Q4_K_M, 512-token prompt and 128 generated tokens, three runs after a warmup, full GPU offload, with power and VRAM sampled throughout. Token generation is memory-bandwidth-bound, so the ranking tracks bandwidth closely, which makes it a fair guide to cards we haven't run yet.