ChatGPT Images 2.0: OpenAI’s new image generation model

OpenAI has launched ChatGPT Images 2.0, the new version of its image generation model, available today for all users of ChatGPT, Codex and the API. The model represents a significant improvement over its predecessor across several technical areas and broadens its potential use in professional contexts.

One of the most notable improvements is greater precision in instruction following. The model can represent objects more accurately, render dense text within images — including icons, interface elements and complex compositions — and produce results at resolutions of up to 2K through the API. It also adds support for aspect ratios ranging from 3:1 to 1:3, making it easier to fit formats such as banners, presentations, posters and social media graphics.

On the language front, the model extends its ability to generate text within images beyond English and Latin-script languages. OpenAI highlights concrete gains in Japanese, Korean, Chinese, Hindi and Bengali, making it easier to create visual materials in the languages users actually work with.

ChatGPT Images 2.0 is also the first OpenAI image model to incorporate reasoning capabilities. When a thinking model is enabled in ChatGPT, the system can search the web for real-time information, generate up to eight distinct images from a single prompt while maintaining visual coherence across them, and review its own outputs before delivery. This functionality is available to users on the Plus, Pro and Business plans.

The model has a knowledge cutoff of December 2025, allowing it to incorporate more up-to-date information when generating informational content such as infographics or educational materials.

Integration with Codex allows images to be generated within the same development environment, without switching tools. Through the API, the model is available under the name gpt-image-2, with pricing that varies depending on the selected quality and resolution.

OpenAI acknowledges limitations in the model: it can struggle with representations that require a complete physical world model, such as origami guides or puzzles, as well as with very dense or repetitive details and diagrams that rely on precise labels or arrows.

Key points

• OpenAI launches ChatGPT Images 2.0 for all users of ChatGPT, Codex and the API.
• Greater precision in instruction following, in-image text and resolutions up to 2K.
• Supports aspect ratios from 3:1 to 1:3 to fit different formats.
• Improved text in non-Latin scripts: Japanese, Korean, Chinese, Hindi and Bengali.
• First OpenAI image model with reasoning capabilities.
• With reasoning enabled, generates up to eight coherent images from a single prompt.
• Advanced features require Plus, Pro or Business plans.
• Available in the API as gpt-image-2, priced by quality and resolution.