Extremely inconsistent vector store indexing times

API Bugs
, ,
1

@OpenAI_Support

I am seeing extreme, intermittent latency in OpenAI-managed vector-store ingestion.

I ran the same capability test twice using the same API project, credentials, endpoint, and ingestion configuration. Each run created a fresh vector store, uploaded an equivalent 145-byte text file, attached it as a one-file batch, and polled the batch status.

Fast run

  • Started: 2026-07-24 15:42:52 UTC
  • Vector-store creation: 0.621s
  • File upload: 1.106s
  • Indexing completed: 6.7s
  • Entire test completed: 20.0s

Slow run

  • Started: 2026-07-24 15:50:52 UTC
  • Vector-store creation: 0.698s
  • File upload: 0.500s
  • Still in_progress with completed=0/1 after 600 seconds
  • Direct vector-store file status remained in_progress
  • last_error remained null

All authentication, permission, file upload, Responses API, and vector-store operations passed. All diagnostic requests returned HTTP 200. There were no 429 responses, rate-limit errors, or 5xx errors. The file itself reached processed; only the asynchronous vector-store ingestion remained stalled.

This appears to isolate the latency to the asynchronous vector-store parsing/chunking/embedding pipeline rather than uploads, permissions, request rate, or file complexity.

Is anyone else observing this degree of ingestion variance? Is there a known ingestion backlog or expected latency range for vector-store file processing?

I have full timestamps, OpenAI request IDs, vector-store IDs, file IDs, batch IDs, and diagnostic output available for a private support case.

2

Thanks for the detailed testing, @Neil_Lewis. There was a degradation posted on the OpenAI Status Page around that period, which may have indirectly affected vector-store file processing. (OpenAI Status)

Are you still seeing the issue? If so, please send the request IDs, UTC timestamps, file IDs, batch IDs, and diagnostic output to support@openai.com, then share the case ID here so the team can help monitor it.

-Mark G.