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/Mistral-Nemo-Instruct-2407-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA B300 | 196.92 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 183.92 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 181.3 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 177.38 tok/s | 80GB | ✓ Measured |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 164.28 tok/s | 96GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 111.07 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 106.33 tok/s | 40GB | Est. |
| NVIDIA L40S | 89.72 tok/s | 48GB | ✓ Measured |
| NVIDIA A10G | 57 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 33.01 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 22.98 tok/s | 16GB | ✓ Measured |
NVIDIA B300 is the fastest card we've measured running Mistral-Nemo-Instruct-2407, at 196.92 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs Mistral-Nemo-Instruct-2407, at 22.98 tok/s.