Skip to content
Updated minutes ago
nvidia

A4000

nvidia · ampere · 16 GB GDDR6 · 140W TDP

VRAM

16 GB

BF16 TFLOPS

76

Bandwidth

448 GB/s

From

$0.17/hr

Calculate ROI with this GPU →

Spec Sheet

VRAM16 GB GDDR6
Memory Bandwidth448 GB/s
BF16 TFLOPS76
FP16 TFLOPS76
FP8 TFLOPS76
INT8 TOPS76
TDP140W
InterconnectPCIE
Max per Node8
PCIe Gen4
CUDA Compute Capability8.6
Tensor CoresYes

Pricing by Provider

ProviderOn-DemandReservedSpotBadge
tensordock$0.25/hr-$0.17/hrCheapest
vast_ai$0.30/hr-$0.18/hr

Compatible Models (246)

Training Capabilities

Estimated GPU count for full fine-tuning (AdamW, BF16) and QLoRA

Model SizeFull Fine-TuneQLoRA
7B model9 GPUs1 GPU
13B model16 GPUs1 GPU
70B model83 GPUs3 GPUs

Energy Efficiency

Estimated tokens/second per Watt for popular models

Mistral 7B
0.44 t/s/WFP8
Qwen 2.5 7B
0.42 t/s/WFP8
Llama 3.1 8B
0.40 t/s/WFP8
Llama 3.1 70B
0.05 t/s/WFP8
Qwen 2.5 72B
0.04 t/s/WFP8

Similar GPUs

GPUVRAMBF16 TFLOPSBW (GB/s)From
A216 GB18200$0.15/hr
RTX 306012 GB12.7360$0.06/hr
RTX 308010 GB47760$0.14/hr
A10G24 GB35600$0.30/hr
A3024 GB165933$0.35/hr

Methodology Note

Performance estimates for the A4000are based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (16 GB GDDR6 available), KV-cache allocation, and activation memory. Throughput predictions use the A4000's rated 448 GB/s memory bandwidth and 76 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (ampere). See our full methodology.

Frequently Asked Questions

How many AI models can run on A4000?

The A4000 can run 246 AI models from our database within a single node. Compatible models range across various parameter sizes depending on the quantization precision (BF16, FP8, INT4). Smaller models fit on a single GPU while larger models may require multi-GPU setups up to 8x A4000.

What is the A4000 inference throughput?

The A4000 delivers 76 BF16 TFLOPS and 76 FP8 TFLOPS with 448 GB/s memory bandwidth. Actual inference throughput (tokens/sec) depends on the model size, precision, and batch size. Use our calculator for model-specific throughput estimates.

How much does A4000 cost per hour?

The A4000 is available starting from $0.17/hour via tensordock. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.