← Workflow
Optional · Step 0 · Quick check
KV Cache Calculator
Estimate KV cache memory from model, context, batch, and precision.
Results
- KV bytes / token
- 327680 B
- KV cache total
- 21.47 GB
- Per GPU (even split)
- 21.47 GB
- GPU capacity
- 80 GB
Optional step 0 — continue to Define workload (Lens) for the full platform workflow.