OpenAI announced ChatGPT Images 2.0 on 21 April 2026, marking a significant update to its image generation capabilities within ChatGPT [source]. The release introduces what the company describes as improved visual quality, faster generation speeds, and enhanced prompt adherence compared to the previous DALL·E integration.

According to the announcement, Images 2.0 processes requests approximately 40% faster than its predecessor while maintaining higher resolution output. OpenAI states the model demonstrates better understanding of spatial relationships, lighting conditions, and stylistic instructions. The update applies to ChatGPT Plus, Team, and Enterprise subscribers, with availability rolling out over several days.

The company disclosed that Images 2.0 incorporates updated safety filters designed to reject requests for public figure likenesses, copyrighted characters, and graphic content. OpenAI notes these filters may occasionally produce false positives, blocking legitimate requests. Users can appeal blocked generations through an in-app feedback mechanism.

OpenAI confirmed that images generated by the new system include C2PA metadata and visible watermarking to indicate AI provenance. The announcement specifies that watermarks appear as subtle patterns embedded during generation, detectable by compatible verification tools.

The rollout follows OpenAI's pattern of incremental capability releases. The company provided no technical details about the underlying architecture or training data composition. No information was shared regarding whether Images 2.0 represents a new model family or an iteration of existing DALL·E technology.

ChatGPT Free tier users retain access to a limited number of image generations per day, though OpenAI did not specify whether these users receive the 2.0 upgrade immediately or on a delayed schedule.