docs/macros/platform-gpu-table.md
| GPU | Generation | VRAM | Cost/Hour | Best For |
|---|---|---|---|---|
| RTX 2000 Ada | Ada | 16 GB | $0.24 | Small datasets, testing |
| RTX A4500 | Ampere | 20 GB | $0.25 | Small-medium datasets |
| RTX 4000 Ada | Ada | 20 GB | $0.26 | Medium datasets |
| RTX A5000 | Ampere | 24 GB | $0.27 | Medium datasets |
| L4 | Ada | 24 GB | $0.39 | Inference optimized |
| A40 | Ampere | 48 GB | $0.40 | Larger batch sizes |
| RTX 3090 | Ampere | 24 GB | $0.46 | General training |
| RTX A6000 | Ampere | 48 GB | $0.49 | Large models |
| RTX PRO 4500 | Blackwell | 32 GB | $0.54 | Great price/performance |
| RTX 4090 | Ada | 24 GB | $0.59 | Best price/performance |
| RTX 6000 Ada | Ada | 48 GB | $0.77 | Large batch training |
| L40S | Ada | 48 GB | $0.86 | Large batch training |
| RTX 5090 | Blackwell | 32 GB | $0.89 | Latest consumer generation |
| L40 | Ada | 48 GB | $0.99 | Large models |
| A100 PCIe | Ampere | 80 GB | $1.39 | Production training |
| A100 SXM | Ampere | 80 GB | $1.49 | Production training |
| RTX PRO 6000 | Blackwell | 96 GB | $1.69 | Recommended default |
| H100 PCIe | Hopper | 80 GB | $2.39 | High-performance training |
| H100 SXM | Hopper | 80 GB | $2.69 | Fastest training |
| H100 NVL | Hopper | 94 GB | $3.07 | Maximum performance |
| H200 NVL | Hopper | 143 GB | $3.39 | Maximum memory (Pro+) |
| H200 SXM | Hopper | 141 GB | $3.59 | Maximum performance (Pro+) |
| B200 | Blackwell | 180 GB | $4.99 | Largest models (Pro+) |