Fastest GPU For Vision Language

Describing and reasoning about a picture: captioning, OCR and visual question answering.

Florence-2 Base

Weights: microsoft/Florence-2-base

NVIDIA H100 80GB HBM3249.62
NVIDIA A10G210.79
NVIDIA L40S184.2
NVIDIA L4174.65
NVIDIA T4148.61
NVIDIA A100 80GB SXM4126.4

Measured in images/min. Longer is faster.

#GPUimages/min
1NVIDIA H100 80GB HBM3249.62
2NVIDIA A10G210.79
3NVIDIA L40S184.2
4NVIDIA L4174.65
5NVIDIA T4148.61
6NVIDIA A100 80GB SXM4126.4

Florence-2 Large

Weights: microsoft/Florence-2-large

NVIDIA H100 80GB HBM3156.15
NVIDIA A10G117.94
NVIDIA L40S106.54
NVIDIA L493.77
NVIDIA T483.43
NVIDIA A100 80GB SXM470.94

Measured in images/min. Longer is faster.

#GPUimages/min
1NVIDIA H100 80GB HBM3156.15
2NVIDIA A10G117.94
3NVIDIA L40S106.54
4NVIDIA L493.77
5NVIDIA T483.43
6NVIDIA A100 80GB SXM470.94

Every number is a first-party run on our own bench. Cards missing from a board have not been measured on that model yet, or cannot fit it.