OpenAI media model shutdown calendar 2026: DALL-E, Sora, GPT Image

OpenAI's deprecations page lists three media retirements for 2026: DALL-E 2 and 3 on May 12, Sora 2 and the Videos API on September 24, and gpt-image-1.5, gpt-image-1-mini and chatgpt-image-latest on December 1. Two have passed, one is still ahead, and the only vendor-named replacements are GPT Image models.

What are the dates?

Everything below comes from the deprecations page read on 2026-10-02.

What is already broken versus still working?

DALL-E and Sora are gone, so any call to them fails today. The December 1 group still works as of this date, which gives you about two months. Note that DALL-E's own suggested replacement list includes gpt-image-1-mini, which is itself on the December 1 list, so do not migrate DALL-E code to a model that is about to go.

What does the image guide recommend now?

OpenAI's current image generation guide lists gpt-image-2.5-sunburst, described as editing precision focused, and gpt-image-2.5-flare, described as fast, high-quality everyday use. Both take low, medium, high, xhigh, max or auto. The Flare page lists the dated snapshot gpt-image-2.5-flare-2026-09-08.

What does Sume do?

Sume's Image API serves the 2.5 models as openai/gpt-image-2.5 (Flare) and openai/gpt-image-2.5-sunburst, and says ChatGPT Image 2 remains selectable. Legacy bare ids such as gpt-image-2 are accepted as aliases for their org/slug forms. For video, there is no Sora path; use the video catalog.

Action list: search your repo for the six retired ids, move still images to a 2.5 id, move video to a catalog id, and set the model explicitly instead of relying on chatgpt-image-latest. Then verify with one real request each.

More detail on the December cutover is in the chatgpt-image-latest post.