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/EVA-Qwen2.5-14B-v0.2-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA B300 | 159.46 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 150.67 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 148.74 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 145.15 tok/s | 80GB | ✓ Measured |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 135.83 tok/s | 96GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 90.35 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 86.48 tok/s | 40GB | Est. |
| NVIDIA L40S | 74.63 tok/s | 48GB | ✓ Measured |
| NVIDIA A10G | 47 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 27.3 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 19.92 tok/s | 16GB | ✓ Measured |
NVIDIA B300 is the fastest card we've measured running EVA-Qwen2.5-14B-v0.2, at 159.46 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs EVA-Qwen2.5-14B-v0.2, at 19.92 tok/s.