Qwen2.5 7B served needs roughly 20GB of VRAM at our tested precision. 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: Qwen/Qwen2.5-7B
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA B200 | 5799.1 serve tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 4394.8 serve tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 3601.1 serve tok/s | 80GB | ✓ Measured |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 2282.8 serve tok/s | 96GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 2187.2 serve tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 1769.6 serve tok/s | 40GB | Est. |
| NVIDIA L40S | 1406.8 serve tok/s | 48GB | ✓ Measured |
| NVIDIA A10G | 858.7 serve tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 469.9 serve tok/s | 24GB | ✓ Measured |
These cards' VRAM is below the requirement at tested precision, published as data, not hidden.
| GPU | VRAM |
|---|---|
| NVIDIA T4 | 16GB |
Qwen2.5 7B served needs roughly 20GB of VRAM at our tested precision. 9 of the GPUs on our bench run it; 1 don't have the VRAM for it.
NVIDIA B200 is the fastest card we've measured running Qwen2.5 7B served, at 5799.1 serve tok/s.
NVIDIA L4 ($2,500 MSRP) is the cheapest card on our bench that runs Qwen2.5 7B served, at 469.9 serve tok/s.