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/mistralai_Mistral-Small-3.1-24B-Instruct-2503-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 108.88 tok/s | 96GB | ✓ Measured |
| NVIDIA B300 | 108.37 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 87.72 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 85.1 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 84 tok/s | 80GB | ✓ Measured |
| NVIDIA L40S | 58.21 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 46.63 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 45.34 tok/s | 40GB | Est. |
| NVIDIA A10G | 26.8 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 16.62 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 10.13 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running Mistral Small 24B (Q3_K_M), at 108.88 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs Mistral Small 24B (Q3_K_M), at 10.13 tok/s.