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: unsloth/LFM2.5-1.2B-Instruct-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 989.91 tok/s | 96GB | ✓ Measured |
| NVIDIA B300 | 955.13 tok/s | 288GB | ✓ Measured |
| NVIDIA H200 | 946.53 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 926.61 tok/s | 80GB | ✓ Measured |
| NVIDIA B200 | 916.79 tok/s | 192GB | ✓ Measured |
| NVIDIA L40S | 661.54 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 576.95 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 572.69 tok/s | 40GB | Est. |
| NVIDIA A10G | 429.96 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 278.53 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 221.66 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running LFM2.5-1.2B, at 989.91 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs LFM2.5-1.2B, at 221.66 tok/s.