Module 1 · Spending less on the same work · Lesson 4
Batch, flex and fast tiers
Paying half price for work nobody is waiting for and a premium for work someone is, which workflow steps belong in which tier, and how to run batch jobs durably without breaking caching or data-handling commitments.
By the end of this lesson you can
- Compare batch, flex, standard and fast or priority tiers across providers on price, latency, failure mode and data handling
- Classify each step of an agent pipeline by latency tolerance and explain why agent loops rarely fit batch
- Compute the saving from batching a bulk job, including the hit rate below which adding a 1-hour cache makes it dearer
- Implement an idempotent, resumable batch job and handle every result type