← Workflow
Optional · Step 0 · Quick check
Quantization Savings
Compare memory and throughput across precisions.
Precision comparison (H100)
| Precision | Weight GB | Est. t/s |
|---|---|---|
| FP16 | 140.0 | 536 |
| FP8 | 70.0 | 1,072 |
| INT8 | 70.0 | 1,072 |
| INT4 | 35.0 | 2,144 |