Skip to main content
2026/8/26 13:15 (UTC+8) · Model Status · OpenAI ✅ gpt-image-2-vip is back to normal speed — you can push concurrency again Yesterday afternoon’s saturation has cleared. In today’s logs from 13:13–13:14 (UTC+8), time-to-first-byte on non-streaming calls sits mostly at 37–55 seconds, with a few at 66–71 seconds — back to everyday levels after yesterday’s 82–190 seconds. The temporary advice to hold concurrency down and retry 429s yourself no longer needs to be kept; raise concurrency as your workload requires.
Console log list showing several non-streaming gpt-image-2-vip calls with time-to-first-byte ranging from 37 to 71 seconds

gpt-image-2-vip call logs from 13:13–13:14 (UTC+8) on 26 August, time-to-first-byte clustered at 37–55 seconds

The provider has also confirmed that gpt-image-2-vip supports 4K output and the quality parameter again:
  • Endpoints: /v1/images/generations for generation, /v1/images/edits for editing
  • size: the common 1K / 2K / 4K sizes
  • quality: low / medium / high
  • Billing: $0.03 per request, so call it with a per-request billing token in the default group
See the gpt-image-2 overview for parameters and code samples. Thanks for your patience over the past two days — we are keeping an eye on this route’s load.
← Back to Live Updates · 📚 Archive by month