The Batch API is the current darling of the cost-conscious. The arrangement is this: if you are willing to wait up to twenty-four hours for your tokens, the vendor will process them at half price. Patience, finally, rewarded.
We have run this exact program since the invention of the bulk coin purchase. We call it the bucket.
The bucket, explained for newcomers
You approach the counter. You request not one token but a great many. The proprietor, pleased by the volume, applies a discount. You receive a bucket. The tokens do not arrive faster, but they arrive cheaper, and they arrive in your physical possession, which the Batch API has never once managed to do.
The twenty-four-hour wait the AI vendors ask of you is, by coincidence, the exact attention span of the average eight-year-old holding a bucket of tokens. We mention this only because it suggests the model was not new.
Asynchronous, but honest
The Batch API's discount evaporates the moment the tokens are spent, which is immediately, because a generated token has no afterlife. A bucket of brass tokens is asynchronous in a more permanent sense: you may spend them this year, next year, or never, and the discount you received remains baked into every disc.
This catalog holds 846 tokens that were, in their way, all bought in batches, by parents at counters, decades ago. They trade today at an average of $13.52. The batch held its value. The batch always holds its value.
The verdict
A 50% discount on something you keep forever beats a 50% discount on something deleted in a second. For how these batches appreciate rather than vanish, read pricing trends over time, or return to the field guide to tokens.
In brass we trust.