Why does Batch API show more requests than expected?
OpenAI Developer Community
March 12, 2026
This is still an issue. My experience:
On 3/12/2026, I am only running the batches shown in the attached screenshot, which account for a total of 64 requests. However, when I review the logs for that same date, I’m seeing 69 requests — 5 more than expected. The usage report also reflects 69 requests. This discrepancy is troubling because these are just small tests right now, and once I scale up my operations, these kinds of issues could become much more significant. I’ve also noticed that the batch API performance has degraded compared to how it was performing previously.
This is getting a bit concerning. I am seeing increased logs, now at 78. Is the batch api rerunning requests that have already been completed? I think this is an urgent issue that requires addressing from the technical team. Please help with this matter.
Discussion in the ATmosphere