← Workflow
Optional · Step 0 · Quick check
Tokens/sec Estimator
Theoretical decode throughput from roofline model.
Results
- Aggregate throughput
- 2,345 t/s
- Per-request decode
- 73.3 t/s
- ms / token (batch)
- 13.65 ms
Optional step 0 — continue to Define workload (Lens) for the full platform workflow.