OpenAI's DALL-E replacement list names models that retire Dec 1
OpenAI's deprecations page maps dall-e-2 and dall-e-3 to gpt-image models, but gpt-image-1-mini retires Dec 1, 2026. Which ids to pick instead on Sume.

On OpenAI's deprecations page, dall-e-2 and dall-e-3 (removed May 12, 2026) list gpt-image-2, gpt-image-1 or gpt-image-1-mini as replacements, but the same page lists gpt-image-1-mini for removal on December 1, 2026. If you followed that old replacement list, check your model id again. On Sume, the current ids are openai/gpt-image-2.5 (Flare) and openai/gpt-image-2.5-sunburst.
Two entries, two dates
The OpenAI deprecations page has one entry for the DALL-E models and one for older GPT Image models. Reading them together shows a replacement chain that has already shortened.
| Retiring id | Announced | Removal date | Replacement listed |
|---|---|---|---|
| dall-e-2 | 2025-11-14 | 2026-05-12 | gpt-image-2, gpt-image-1 or gpt-image-1-mini |
| dall-e-3 | 2025-11-14 | 2026-05-12 | gpt-image-2, gpt-image-1 or gpt-image-1-mini |
| gpt-image-1-mini | 2026-06-02 | 2026-12-01 | gpt-image-2.5-sunburst or gpt-image-2.5-flare |
| gpt-image-1.5 | 2026-06-02 | 2026-12-01 | gpt-image-2.5-sunburst or gpt-image-2.5-flare |
| chatgpt-image-latest | 2026-06-02 | 2026-12-01 | gpt-image-2.5-sunburst or gpt-image-2.5-flare |
What to take from it
The DALL-E replacement text was written in November 2025. gpt-image-1-mini was a sensible landing spot then; by the June 2026 notice it was already on the removal list. A project that migrated from DALL-E to gpt-image-1-mini has to move a second time before December 1.
The June entry names the current targets: gpt-image-2.5-sunburst and gpt-image-2.5-flare. OpenAI's image generation guide describes Sunburst as optimised for editing precision and Flare as the fast, high-quality everyday model.
Ids on Sume
Sume's Image API docs name openai/gpt-image-2.5 for Flare and openai/gpt-image-2.5-sunburst, and say ChatGPT Image 2 remains selectable. Both 2.5 ids take up to 16 image references, an optional mask_url, and background set to auto, transparent or opaque. An omitted quality defaults to high.
Sume's docs also say bare Image Router ids such as gpt-image-2 are accepted as aliases for their org/slug forms. Use the full org/slug id in new code.
Migration checklist
Replace the id, then check every parameter you send. A request that sets a parameter the chosen model does not list is rejected with 400 unsupported_parameter, not silently dropped; Sume's docs say output_compression and seed are not served in v1. Compare a handful of your real prompts side by side before cutting over, and move the work before December 1 rather than on it. The December 1 migration post has the Sume-side id mapping.
Sources
Related posts
More in Developers
- OpenAI structured outputs: 5000 properties, 10 levels, vs Sume
OpenAI caps strict schemas at 5000 properties and 10 nesting levels. Sume's output_schema uses the same numbers but counts enums and strings differently.
- OpenAI retries webhooks 72 hours; Sume stops after ten attempts
OpenAI retries a failing webhook for up to 72 hours. Sume stops after ten attempts, about 5 minutes for jobs and 3 hours for runs, so keep a status poll.
- OpenRouter output_modalities=video vs Sume /v1/videos/models
OpenRouter lists video models three ways. Sume has a /v1/videos/models endpoint with the same shape plus a catalog. Fields to read before you submit.
- OpenRouter video is ZDR-ineligible: what Sume says about retention
OpenRouter says video generation cannot use Zero Data Retention because output is held for retrieval. Sume has no ZDR toggle; what to tell your security team.
Written by Sume