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/GLM-4.7-Flash-REAP-23B-A3B-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 199.14 tok/s | 96GB | ✓ Measured |
| NVIDIA B300 | 178.49 tok/s | 288GB | ✓ Measured |
| NVIDIA H200 | 171.45 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 168.41 tok/s | 80GB | ✓ Measured |
| NVIDIA B200 | 167.76 tok/s | 192GB | ✓ Measured |
| NVIDIA L40S | 146.51 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 112.75 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 109.1 tok/s | 40GB | Est. |
| NVIDIA A10G | 97.48 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 70.87 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 53.83 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running GLM-4.7-Flash-REAP-23B-A3B, at 199.14 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs GLM-4.7-Flash-REAP-23B-A3B, at 53.83 tok/s.