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/gemma-2-2b-it-abliterated-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA B300 | 418.18 tok/s | 288GB | ✓ Measured |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 416.68 tok/s | 96GB | ✓ Measured |
| NVIDIA H200 | 397.43 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 392.65 tok/s | 80GB | ✓ Measured |
| NVIDIA B200 | 392.32 tok/s | 192GB | ✓ Measured |
| NVIDIA L40S | 278.14 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 244.41 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 232.27 tok/s | 40GB | Est. |
| NVIDIA A10G | 174.56 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 113 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 91.91 tok/s | 16GB | ✓ Measured |
NVIDIA B300 is the fastest card we've measured running gemma-2-2b-it-abliterated, at 418.18 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs gemma-2-2b-it-abliterated, at 91.91 tok/s.