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/DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA B300 | 292.55 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 274.3 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 267.82 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 262.76 tok/s | 80GB | ✓ Measured |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 239.45 tok/s | 96GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 164.46 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 156.62 tok/s | 40GB | Est. |
| NVIDIA L40S | 134.78 tok/s | 48GB | ✓ Measured |
| NVIDIA A10G | 87.84 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 48.82 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 35.73 tok/s | 16GB | ✓ Measured |
NVIDIA B300 is the fastest card we've measured running DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored, at 292.55 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored, at 35.73 tok/s.