OpenAI has released ChatGPT Images 2.0, a new image generation model that is now available across ChatGPT, Codex and its API. It allows users to create more accurate and detailed images for tasks like design, presentations and content creation. According to OpenAI, the new model focuses on making image generation more practical for everyday use, going beyond simple visual experiments. It can also use web data when needed and create multiple outputs from a single prompt.
ChatGPT Images 2.0 is available for all ChatGPT and Codex users. However, advanced features like thinking-enabled outputs are limited to Plus, Pro, Business and Enterprise users.
ChatGPT Images 2.0: What’s new
Improved image generation capabilities
With Images 2.0, OpenAI said that the model is better at understanding detailed instructions and creating images with more precision. It can handle complex elements such as small text, diagrams, UI components and dense layouts more effectively than before. The model is also designed to place objects more accurately within an image and support multiple languages.
Thinking-based workflows
The new update introduces “thinking-enabled” workflows within ChatGPT. This allows the system to reason through a task before generating images. It can also use web data when needed and create multiple outputs from a single prompt.
Users can now generate up to eight related images in one request. This can be useful for creating storyboards, poster sets, manga-style pages or multi-format campaigns without needing separate prompts for each variation.
Developer access through API
OpenAI has also made the underlying model, called gpt-image-2, available through its API. This allows developers to integrate image generation and editing features directly into their apps and workflows.
Safety guidelines
OpenAI said the safety system for Images 2.0 builds on the same foundation as its earlier model, with added safeguards to address risks that may come with more advanced capabilities. According to OpenAI, ChatGPT Images 2.0 can create more realistic images than earlier versions, which could increase the risk of misuse, such as generating misleading or sensitive images of real people, places or events. To prevent this, the company said that it has put strict safety rules in place. The system checks requests before image generation, filters any risky input images and reviews the final output to ensure it follows policies.
Additionally, OpenAI said it tested its safety system using specially designed prompts that try to create content across different risk areas, such as violence or sexual content. These tests are meant to check how well the system can block harmful outputs. The company also clarified that these test scenarios are extreme and do not reflect how often such requests appear in real-world use.
