Aurora alpha models pulled after brief API appearance, now surfacing in Arena image tests
Three leaked Aurora model IDs were pulled hours after surfacing in OpenAI's API, then turned up in Arena blind image tests. Observers guess they could be an early look at GPT-Image-3.
What happened
Three leaked OpenAI model identifiers, aurora-red-alpha, aurora-green-alpha, and aurora-blue-alpha, briefly appeared in OpenAI's developer API this weekend and were removed within hours, which looks like an accidental rollout. The identifiers have since turned up in blind Arena tests, where they appear to be image models: early samples from aurora-green-alpha show clean, detailed output, with the red variant running faster at lower quality and green slower at higher quality. The blue variant has not been tested yet. (Source: TestingCatalog, October 18, 2026.)
Why it matters
This is the first evidence of what the Aurora identifiers actually do, image generation rather than text, and it points at a possible GPT-Image-3 release. Both TestingCatalog and Lumina Bench stress this is still a guess: OpenAI has published no documentation, pricing, or rate limits for any of the three models.
Sources
- https://www.testingcatalog.com/openai-possibly-testing-gpt-image-3/
- https://cryptobriefing.com/openai-aurora-alpha-models-api/
Verified 2026-10-18.
Sources
Get updates like this every morning
- ② Card on Stripe
- ③ 7 days free
Then $2/month · cancel anytime in one click