batchwatch › OpenAI › gpt-5-nano
Measured, not promised. OpenAI publishes a completion window of 24 hours and nothing more precise; below is what 7 real batch jobs actually did.
Share of jobs finished, against elapsed time. 7 completions, 30-day window.
The curve is drawn from the pooled measurements, while the headline percentiles above are the median of each contributor's own figures - so the two need not line up exactly. That is deliberate: no single source can move a headline number by sending more data.
Provider's published window - not a measurement
Plan for 24 hours - and note where that number comes from. With 7 measurements there is no order statistic that can bound the upper end, so the honest planning number is OpenAI's own published completion window. That is a contract they wrote, not something we observed.
For contrast, what we did observe: a p90 of 65 min across 7 jobs. Too thin to plan on, which is exactly why it is not the number above.
Score 0.18 · basis: exact_model
Limited by only 7 measurements, a single contributor, an upper bound we cannot pin down.
Pooled across all measurements: fewer than 3 contributors have enough data to vote, so a single contributor could move this number. Confidence is capped accordingly.
Each measurement is one completed batch job. The duration is end minus start, both timestamped by our server - a client can never submit a duration, and a job we stopped waiting for is recorded as abandoned, never as a fast completion.
| Newest measurement | 2026-08-25T18:25:59.000Z |
|---|---|
| Oldest in window | 2026-08-25T13:39:43.000Z |
| Figures computed | 2026-08-25T18:42:24.000Z |
| Public delay | 15 minutes |
| Percentile method | pooled (1 voting source) |
Timestamps are absolute and in UTC on purpose: this page may be cached, and a relative age would quietly go stale in the cache while an absolute one stays true.
Same numbers, same delay, no key required.
curl "https://batchwatch.dev/v1/curve?provider=openai&model=gpt-5-nano"
Contributing five measurements in seven days cuts the delay to five minutes and unlocks the decision endpoints (/v1/should-i-batch, /v1/estimate-batchtime). Start on the front page.