Story

In brief

Sizing the DSpark draft-verification budget from per-request confidence instead of verifying every drafted token, so one configuration holds the throughput/latency frontier from batch size 1 to 256.