A model that doesn't fit doesn't run slowly, it doesn't run. Here's every card from our bench, sorted by measured speed.
Weights: microsoft/Florence-2-large
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA H100 80GB HBM3 | 156.15 images/min | 80GB | ✓ Measured |
| NVIDIA A10G | 117.94 images/min | 24GB | ✓ Measured |
| NVIDIA L40S | 106.54 images/min | 48GB | ✓ Measured |
| NVIDIA L4 | 93.77 images/min | 24GB | ✓ Measured |
| NVIDIA T4 | 83.43 images/min | 16GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 70.94 images/min | 80GB | ✓ Measured |
NVIDIA H100 80GB HBM3 is the fastest card we've measured running Florence-2 Large, at 156.15 images/min.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs Florence-2 Large, at 83.43 images/min.