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/Qwen_Qwen3-1.7B-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 593.5 tok/s | 96GB | ✓ Measured |
| NVIDIA B300 | 575.97 tok/s | 288GB | ✓ Measured |
| NVIDIA H200 | 568.85 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 557.09 tok/s | 80GB | ✓ Measured |
| NVIDIA B200 | 551.93 tok/s | 192GB | ✓ Measured |
| NVIDIA L40S | 405.3 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 350.4 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 341.85 tok/s | 40GB | Est. |
| NVIDIA A10G | 262.49 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 176.08 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 135.21 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running Qwen3-1.7B, at 593.5 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs Qwen3-1.7B, at 135.21 tok/s.