How long does the OpenAI batch API actually take?

The provider quotes "within 24 hours", but that is a ceiling, not the typical time. Measured, the median is far below it: gpt-5.6-sol clears in 78s right now. The exact number moves with OpenAI's backlog and differs by model, which is why we render it live rather than quoting a fixed figure.

Is the OpenAI batch API worth using?

For work that can wait, almost always: it costs 50% of the synchronous rate - half price on both input and output tokens - and the measured wait is typically minutes, not the "24 hours" the ceiling implies. Whether it fits depends on your deadline, not on folklore. Send your limit to /v1/should-i-batch and we answer against it.

Which OpenAI model has the fastest batch queue?

The queue is not one queue: OpenAI's models sit in different backlogs, and the gap between the fastest and slowest we can quote is often wider than the gap between providers. The table above ranks every model we measure, fastest median first, updated live.

Are these numbers measured or the provider's window?

Measured. Each is an observed completion of a real batch job, timed on our server from submit to result - not OpenAI's "up to 24 hours" marketing window. We publish a median only once a model has at least 8 completions; below that we show the count and say so, because a median on a handful of jobs is just the slowest of the handful.

OpenAI batch API: how long does it take?

The measured answer

Right now the fastest OpenAI model clears in a measured median of 78s, the slowest we can quote in 5 min - the queue is not one queue, and the model you pick matters more than the provider.

Every figure here is an observed completion of a real batch job, timed on our server from the moment we submit to the moment the result comes back - not OpenAI's published "within 24 hours" window, which is the ceiling everyone quotes and nobody measures. The queue moves with OpenAI's backlog, so we render the number live rather than freezing a figure that would be wrong within days.

Every OpenAI model, measured

Model	Median wait (p50)	p90	n
gpt-5.6-sol	78s	10 min	520
gpt-5-nano	88s	3 hours	508
gpt-5.6-luna	5 min	5 hours	561

Rendered live from the rollup; newest measurement 2026-08-30, from 1,589 measured OpenAI batch jobs.

Half the price - for a wait you can now see

OpenAI's batch API costs 50% of the synchronous rate on both input and output tokens - that is OpenAI's own published price, not a batchwatch estimate. The saving is flat; the wait is what varies, and it is the number this page exists to give you straight. Weigh one against the other with real figures instead of "up to 24 hours". See the full price table paired with each model's measured wait.

Questions this answers

How long does the OpenAI batch API actually take?: The provider quotes "within 24 hours", but that is a ceiling, not the typical time. Measured, the median is far below it: gpt-5.6-sol clears in 78s right now. The exact number moves with OpenAI's backlog and differs by model, which is why we render it live rather than quoting a fixed figure.
Is the OpenAI batch API worth using?: For work that can wait, almost always: it costs 50% of the synchronous rate - half price on both input and output tokens - and the measured wait is typically minutes, not the "24 hours" the ceiling implies. Whether it fits depends on your deadline, not on folklore. Send your limit to /v1/should-i-batch and we answer against it.
Which OpenAI model has the fastest batch queue?: The queue is not one queue: OpenAI's models sit in different backlogs, and the gap between the fastest and slowest we can quote is often wider than the gap between providers. The table above ranks every model we measure, fastest median first, updated live.
Are these numbers measured or the provider's window?: Measured. Each is an observed completion of a real batch job, timed on our server from submit to result - not OpenAI's "up to 24 hours" marketing window. We publish a median only once a model has at least 8 completions; below that we show the count and say so, because a median on a handful of jobs is just the slowest of the handful.

See it live

Every number here is measured, and it is all public.

The live dashboard shows every model with its full distribution, n and confidence grade. /compare ranks every provider's batch API side by side, and /v1/coverage is the same data as JSON, no key required. Send your own deadline to /v1/should-i-batch and we tell you whether OpenAI's queue meets it.