OpenAI officially retired DALL-E 2 and DALL-E 3 on May 12, 2026, ending the era of the DALL-E product line. Both models are replaced by gpt-image-2, a unified image generation model that powers ChatGPT Images 2.0 and is available via the OpenAI API.
What gpt-image-2 adds
The new model delivers native 2K resolution output without upscaling — a significant quality jump over DALL-E 3's maximum of 1792×1024 pixels. Multilingual text rendering allows gpt-image-2 to accurately generate readable text in non-Latin scripts including Arabic, Chinese, Hebrew, and Cyrillic — a capability DALL-E 3 handled poorly. In thinking mode, the model supports up to 8 images per prompt, enabling rapid variation generation for design and marketing workflows.
API and developer impact
Existing API integrations using DALL-E 3 endpoints require migration to the gpt-image-2 endpoint. OpenAI is providing a compatibility period for existing API customers through Q3 2026. The new model is already available in the ChatGPT interface for Plus and Pro subscribers. API pricing for gpt-image-2 is per-image at resolutions from 512px to 2K, with bulk pricing for high-volume API users.
Why it matters
The retirement of the DALL-E brand after four years signals OpenAI's shift toward a unified model naming convention where image generation is a capability of the GPT model family rather than a separate product line. For users and developers, the immediate benefit is substantially better image quality and text handling — particularly valuable for localized marketing content that requires non-Latin text.