GPT-4o Image Generation
2025-03-25 · Publisher : OpenAI · Domain : Artificial Intelligence
OpenAI builds native image generation into GPT-4o in ChatGPT: a multimodal model that renders text accurately, follows detailed instructions and refines its images over the course of the conversation.
What changes
- Image generation becomes a native capability of GPT-4o, rather than a separate model like DALL·E.
- More reliable text rendering in the image, adherence to complex instructions and multi-turn edits within a conversation.
- Becomes the default generator in ChatGPT for Plus, Pro, Team and Free accounts; also available in Sora. DALL·E remains accessible through a dedicated GPT.
- Each image carries C2PA metadata indicating its origin, supplemented by an internal search tool.
- API access arrives on April 23, 2025 under the name gpt-image-1 (dedicated entry).
### Specifications - Model: GPT-4o (built-in image generation); API identifier not available at launch. - Gradual rollout starting on the day of the announcement; Enterprise and Edu afterwards. - Context window and knowledge cutoff: not disclosed for this capability. ### Pricing Pricing not disclosed at launch (included in ChatGPT subscriptions; API announced for the following weeks). ### Benchmarks No numerical score published by the vendor in the announcement, which relies on visual examples and safety evaluations. ### Key takeaways The launch made image generation a common use of ChatGPT, with copyright and provenance issues that the vendor addresses through C2PA. The original page has since been updated (mention of a later version of the image product).
Links
Tags : Image, OpenAI, GPT-4o, Multimodal, ChatGPT