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/SmolLM2-135M-Instruct-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 1252.77 tok/s | 96GB | ✓ Measured |
| NVIDIA L40S | 916.56 tok/s | 48GB | ✓ Measured |
| NVIDIA H200 | 905.38 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 898.53 tok/s | 80GB | ✓ Measured |
| NVIDIA B300 | 855.87 tok/s | 288GB | ✓ Measured |
| NVIDIA B200 | 815.3 tok/s | 192GB | ✓ Measured |
| NVIDIA A10G | 680.37 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 652.15 tok/s | 24GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 554.44 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 544.06 tok/s | 40GB | Est. |
| NVIDIA T4 | 416.24 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running SmolLM2-135M, at 1252.77 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs SmolLM2-135M, at 416.24 tok/s.