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: bartowski/Qwen2.5-7B-Instruct-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA B300 | 293.63 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 277.17 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 265.15 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 264.17 tok/s | 80GB | ✓ Measured |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 256.63 tok/s | 96GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 166.92 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 159.81 tok/s | 40GB | Est. |
| NVIDIA L40S | 143.94 tok/s | 48GB | ✓ Measured |
| NVIDIA A10G | 89.68 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 53.32 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 39.01 tok/s | 16GB | ✓ Measured |
NVIDIA B300 is the fastest card we've measured running Qwen2.5-7B, at 293.63 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs Qwen2.5-7B, at 39.01 tok/s.