Best Cloud GPUs for Fine-Tuning — May 2026
Fine-tuning means single-GPU or small-cluster training of pre-trained models. Pick a GPU with enough VRAM for your model size and decent FP16 throughput.
NVIDIA
80 GB
A100 SXM (80GB)
HBM2e
Ampere
$1.10/hr
NVIDIA
48 GB
L40S
GDDR6
Ada Lovelace
$0.55/hr
NVIDIA
40 GB
A100 SXM (40GB)
HBM2e
Ampere
$0.80/hr
NVIDIA
96 GB
RTX PRO 6000
GDDR7
Blackwell
$1.71/hr
NVIDIA
48 GB
RTX 6000 Ada
GDDR6
Ada Lovelace
$0.47/hr
NVIDIA
32 GB
RTX 5090
GDDR7
Blackwell
$0.34/hr
NVIDIA
24 GB
RTX 4090
GDDR6X
Ada Lovelace
$0.28/hr
NVIDIA
24 GB
RTX 3090 Ti
GDDR6X
Ampere
NVIDIA
16 GB
RTX 5080
GDDR7
Blackwell
NVIDIA
16 GB
RTX 4080 SUPER
GDDR6X
Ada Lovelace
NVIDIA
16 GB
RTX 4080
GDDR6X
Ada Lovelace
A100 SXM (80GB) vs L40S vs A100 SXM (40GB) — top picks from this guide
|
A100 SXM (80GB)
Ampere · 80 GB
|
L40S
Ada Lovelace · 48 GB
|
A100 SXM (40GB)
Ampere · 40 GB
|
|
|---|---|---|---|
| Specifications | |||
| Manufacturer | NVIDIA | NVIDIA | NVIDIA |
| Architecture | Ampere | Ada Lovelace | Ampere |
| VRAM | 80 GB HBM2e | 48 GB GDDR6 | 40 GB HBM2e |
| Memory Bandwidth | 2,039 GB/s | 864 GB/s | 1,555 GB/s |
| FP16 (Tensor) | 312 TFLOPS | 366 TFLOPS | 312 TFLOPS |
| FP32 | 19.5 TFLOPS | 91.6 TFLOPS | 19.5 TFLOPS |
| TDP | 400 W | 350 W | 400 W |
| Release Year | 2020 | 2023 | 2020 |
| Segment | Data center | Data center | Data center |
| Cloud Pricing | |||
| Cheapest On-Demand | $1.10/hr | $0.55/hr | $0.80/hr |
| Providers | 6 | 7 | 2 |
Build your own GPU comparison
Select any 2 GPUs from this guide and open them side-by-side.
Tip: GPU comparisons run in pairs. Pick exactly 2 — if you skip selection, we open the top 2 from this guide.