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/DeepSeek-R1-Distill-Qwen-14B-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 149.81 tok/s | 96GB | ✓ Measured |
| NVIDIA B300 | 144.74 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 120.47 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 119.66 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 118.43 tok/s | 80GB | ✓ Measured |
| NVIDIA L40S | 85.85 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 67.13 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 65.33 tok/s | 40GB | Est. |
| NVIDIA A10G | 41.72 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 27.87 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 16.86 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running DeepSeek-R1 Distill 14B (Q3_K_M), at 149.81 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs DeepSeek-R1 Distill 14B (Q3_K_M), at 16.86 tok/s.