NVIDIA RTX A4000, AI & Machine Learning Benchmarks & Specs

16GB · AI Score 4.1/100 · first-party measured on 12 AI workloads

4.1 AI Score Includes estimates

Every number on this page is first-party: NVIDIA RTX A4000 was run on our pinned 12-workload AI suite on 2026-07-11, with under 0.5% run-to-run variance. On Llama 3.1 8B (Q4_K_M) NVIDIA RTX A4000 delivers about 75.52 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 16GB. For image generation, SDXL runs at 2.62 it/s, while FLUX.1-dev won't fit at BF16 (needs ~26GB). 6 of the 12 workloads won't fit on 16GB at the tested precision, Qwen3 32B, Llama 3.3 70B, FLUX.1-dev, FLUX.1 Kontext and others. We publish those as hard gates rather than quietly dropping to a smaller quant. NVIDIA RTX A4000 isn't a retail purchase for most people. It's rented by the hour. You can run this exact card on RunPod.

AI & Machine Learning benchmark results

Text Generation tok/s 19

MiniCPM5 2B188.85

LFM2.5 2.6B184.62

Granite 4.1 3B133.61

Nemotron 3 Nano 4B117.86

Qwen3 4B116.23

DeepSeek Coder 7B Instruct v1.586.04

Llama 3 8B76.23

Llama 3.1 8B75.52

Qwen3 8B74.04

Ornith 1.5 9B66.27

Nemotron Nano 9B v254.69

Gemma 4 12B47.14

Workload	Result	Telemetry	Data
MiniCPM5 2B	188.85 tok/s	97 W67°CQ4_K_M	✓ Measured
LFM2.5 2.6B	184.62 tok/s	98 W64°CQ4_K_M	✓ Measured
Granite 4.1 3B	133.61 tok/s	116 W72°CQ4_K_M	✓ Measured
Nemotron 3 Nano 4B	117.86 tok/s	116 W70°CQ4_K_M	✓ Measured
Qwen3 4B	116.23 tok/s	2.9 GB peak73 W49°C1.6 tok/WQ4_K_M	✓ Measured
DeepSeek Coder 7B Instruct v1.5	86.04 tok/s	123 W72°CQ4_K_M	✓ Measured
Llama 3 8B	76.23 tok/s	121 W72°CQ4_K_M	✓ Measured
Llama 3.1 8B	75.52 tok/s	4.8 GB peak84 W59°C0.9 tok/WQ4_K_M	✓ Measured
Qwen3 8B	74.04 tok/s	123 W72°CQ4_K_M	✓ Measured
Nemotron Nano 9B v2	54.69 tok/s	123 W73°CQ4_K_M	✓ Measured
Ornith 1.5 9B	66.27 tok/s	124 W72°CQ4_K_M	✓ Measured
Gemma 4 12B	47.14 tok/s	124 W74°CQ4_K_M	✓ Measured
Qwen2.5-Coder 14B	41.13 tok/s	8.6 GB peak74 W63°C0.55 tok/WQ4_K_M	✓ Measured
Qwen3 14B	42.27 tok/s	127 W74°CQ4_K_M	✓ Measured
Gemma 4 26B A4B	✕ Won't fit needs ~20 GB	VRAM-gated at this precision	Est.
Qwen3 30B A3B	✕ Won't fit needs ~20 GB	VRAM-gated at this precision	Est.
Gemma 4 31B	✕ Won't fit needs ~22 GB	VRAM-gated at this precision	Est.
Qwen3 32B	✕ Won't fit needs ~23 GB	VRAM-gated at this precision	✓ Measured
Llama 3.3 70B	✕ Won't fit needs ~46 GB	VRAM-gated at this precision	✓ Measured

Image Generation images/min 6

Stable Diffusion 1.529.26

Sana 1.6B14.18

PixArt-Sigma XL8.05

Stable Diffusion XL5.24

Playground v2.53.21

Workload	Result	Telemetry	Data
Stable Diffusion 1.5	29.26 images/min	138 W71°C	✓ Measured
Sana 1.6B	14.18 images/min	139 W79°C	✓ Measured
Stable Diffusion XL	5.24 images/min	15.7 GB peak139 W66°C11.5 s/img	✓ Measured
Playground v2.5	3.21 images/min	140 W93°C	✓ Measured
PixArt-Sigma XL	8.05 images/min	140 W86°C	✓ Measured
FLUX.1 dev	✕ Won't fit needs ~26 GB	VRAM-gated at this precision	✓ Measured

Image Editing images/min 2

Workload	Result	Telemetry	Data
FLUX.1 Kontext dev	✕ Won't fit needs ~26 GB	VRAM-gated at this precision	✓ Measured
Qwen-Image-Edit	✕ Won't fit needs ~42 GB	VRAM-gated at this precision	✓ Measured

Video Generation frames/s 2

Workload	Result	Telemetry	Data
LTX-Video (distilled)	2.32 frames/s	128 W92°C41.8 s/clip	✓ Measured CPU offload
Wan 2.2 5B (720p)	✕ Won't fit needs ~18 GB	VRAM-gated at this precision	✓ Measured

Image to 3D assets/hour 1

Workload	Result	Telemetry	Data
TRELLIS.2 Image-to-3D	✕ Won't fit needs ~24 GB	VRAM-gated at this precision	✓ Measured

How we measured this. Every result comes from our own pinned, reproducible AI suite, 12 workloads: the Qwen3-4B to Llama-70B LLM ladder (llama.cpp, Q4_K_M), SDXL / Z-Image / FLUX-dev generation, FLUX-Kontext / Qwen-Edit editing, and LTX / Wan video, run first-party on rented hardware with under 0.5% run-to-run variance. Peak VRAM, power draw, temperature and tokens-per-watt are captured per workload. “Won’t fit” rows are real data: where a model exceeds the card’s VRAM at the tested precision we record a hard gate rather than silently dropping to a smaller quant. Measured 2026-07-11 · harness 2.0.0-standalone.

NVIDIA RTX A4000 specifications

Architecture	Ampere
CUDA cores	6,144
VRAM	16GB GDDR6
Memory bus	256-bit
Memory bandwidth	448 GB/s
Boost clock	1,560 MHz
TDP	140 W
Process	8nm
Interface	PCIe 4.0 x16
Release date	2021-04-12
Launch MSRP	$999

Verdict, capable, but 16GB sets the ceiling

NVIDIA RTX A4000 scores 4.1/100, #48 of 102. It ran 4 of 12; 6 exceeded its 16GB. Every figure here is our own measurement.

Relative performance: where the NVIDIA RTX A4000 lands

100% = this card, AI & Machine Learning headline metric (AI Score). #14 of 20 workstation cards in this vertical.

GPU	%	AI Score
NVIDIA RTX A5000	180%	7.4
NVIDIA RTX 4500 Ada Generation	144%	5.9
NVIDIA RTX A4500	124%	5.1
AMD Radeon Pro W7900	110%	4.5
NVIDIA RTX A4000	100%	4.1
NVIDIA Quadro RTX 5000	85%	3.5
NVIDIA RTX 2000 Ada Generation	76%	3.1
AMD Radeon Pro W6800	73%	3
AMD Radeon Pro W7800	73%	3

One tier up: AMD Radeon Pro W7900 (+10%)
One tier down: NVIDIA Quadro RTX 5000 (-15%) · compare head-to-head

← All AI & Machine Learning GPU rankings

The silicon

Transistors	17,400 million
Die size	392.5 mm²
Process node	8 nm
Fabricated by	Samsung
Transistor density	44.3 million per mm²

Denser than 80% of the 746 cards we have silicon data for. Density is the clearest measure of what a process node bought: a card that gained it without growing the die got its speed from the fab rather than the architecture.

Silicon figures from Wikipedia (CC BY-SA 4.0). Benchmarks on this page are our own. Compare every chip.

What this card can build

Whole-job timings, composed from our measured per-model results on this card.

Workflow	Time	Energy	Basis
Full codebase review	24.3 min	30.11 Wh	measured

Can't run: Product photo shoot (needs FLUX.1 Kontext dev), Photo restoration batch (needs FLUX.1 Kontext dev), Restore and enlarge photos (needs FLUX.1 Kontext dev), Character sheet, 12 poses (needs FLUX.1 dev), Short social clips (needs Qwen3 32B), 60-second AI short film (needs Qwen3 32B), 24-frame storyboard (needs Qwen3 32B), 6-panel comic page (needs Qwen3 32B), Long-form article batch (needs Llama 3.3 70B).

Rent or buy?

This card is $999 to buy. The cheapest listed rate on Vast.ai is $0.088/hour, but that is the floor: we budget $0.106/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 9,460 GPU-hours. Below it you are paying for idle silicon.

How you would use it	GPU-hours a year	Rental cost a year	Time to break even
2 hours a day, hobby	730	$77	13.0 years
8 hours a day, working on it	2,920	$308	3.2 years
24/7, always-on agent	8,760	$925	1.1 years

At hobby usage this card is very unlikely to pay for itself before it is superseded. Rent it. Rental figures include a 20% premium over the cheapest listed rate. Ignores electricity, resale and the fact that a rented card can be a newer one tomorrow.

Rental price

$0.088/hr-48.2% since 2026-08-14low $0.078 · high $0.170

Cheapest of the RunPod and Vast on-demand rates we see, sampled daily. Spot and interruptible pricing runs lower.