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: lmstudio-community/gemma-3-1b-it-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 567.1 tok/s | 96GB | ✓ Measured |
| NVIDIA B300 | 533.32 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 519.19 tok/s | 192GB | ✓ Measured |
| NVIDIA H200 | 513.58 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 504.17 tok/s | 80GB | ✓ Measured |
| NVIDIA L40S | 426.49 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 324.08 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 310.69 tok/s | 40GB | Est. |
| NVIDIA A10G | 284.78 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 216.27 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 156.88 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running gemma-3-1b, at 567.1 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs gemma-3-1b, at 156.88 tok/s.