Batchwatch measures the part of LLM pricing nobody can plan around

Field notes

Batchwatch is organized around measurement jobs sent to the batch endpoints of three major model providers, rather than around generic queue management. Its public calculator separates interactive work from eligible nightly or backfill workloads and estimates savings from the stated 50% batch discount while accounting for missed deadlines. The homepage offers an API key through a one-field flow and links to clients on GitHub, although several displayed speed figures remain marked “measuring now.”