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/Goekdeniz-Guelmez_Josiefied-Qwen3-8B-abliterated-v1-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA B300 | 270.09 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 253.28 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 250.88 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 245 tok/s | 80GB | ✓ Measured |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 227.15 tok/s | 96GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 152.29 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 146.8 tok/s | 40GB | Est. |
| NVIDIA L40S | 130.25 tok/s | 48GB | ✓ Measured |
| NVIDIA A10G | 83.26 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 48.72 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 35.62 tok/s | 16GB | ✓ Measured |
NVIDIA B300 is the fastest card we've measured running Josiefied-Qwen3-8B-abliterated-v1, at 270.09 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs Josiefied-Qwen3-8B-abliterated-v1, at 35.62 tok/s.