OpenAI Unveils ChatGPT Images 2.0 With Improved Image Generation, Reasoning Capabilities

OpenAI on Tuesday launched its next-generation image generation model. Dubbed ChatGPT Images 2.0, it is claimed to deliver more precise, usable, and context-aware images, based on prompts entered by the user. The new model introduces improvements in instruction following, multilingual rendering, and composition. The San Francisco-based artificial intelligence (AI) giant says it also adds reasoning capabilities for more complex tasks. ChatGPT Images 2.0 is being rolled out across ChatGPT, Codex, and the API.

Images 2.0 Availability

ChatGPT Images 2.0 is available starting Tuesday to all users across ChatGPT and Codex, as per OpenAI. Advanced capabilities, including reasoning-backed thinking features, are accessible to ChatGPT Plus, Pro, and Business subscribers.

Developers can access the model through the gpt-image-2 API. Pricing, however, will vary based on the selected image quality and resolution. The company claims its new model supports outputs of up to 2K resolution, although higher-resolution outputs are still in the beta phase.

ChatGPT Images 2.0 Features, Capabilities

According to OpenAI, ChatGPT Images 2.0 offers greater precision and control than the previous-generation model, which allows users to generate visuals that closely match detailed prompts. It is claimed to be designed to handle complex compositions. This includes UI elements, dense text, and structured layouts, which the AI firm says were previously challenging for image generation systems.

A notable upgrade is in terms of multilingual support. It can render non-English text more accurately, covering languages like Hindi, Bengali, Chinese, Japanese, and Korean. Consequently, users can create visuals like posters, diagrams, and infographics in the aforementioned languages, where language is an integral part of the design.

As per the company, ChatGPT Images 2.0 also delivers better consistency across styles, including photorealistic images, cinematic visuals, pixel art, and manga. There are improvements on the lighting, texture, and composition fronts, too. The latest image-generation model also supports flexible aspect ratios, from ultra-wide formats (3:1) to tall layouts (1:3).

With ChatGPT Images 2.0, OpenAI has also introduced thinking capabilities. As described by the AI firm, this allows the model to carry out more advanced reasoning tasks than before. It can be paired with a reasoning-enabled ChatGPT model, enabling it to search the web for real-time information, verify outputs, and generate images — all from a single prompt. It has an updated knowledge cutoff of December 2025.

OpenAI claims the model can generate up to eight coherent outputs simultaneously, while maintaining consistency across characters, objects, and other elements. The company has positioned ChatGPT Images 2.0 for a wide range of use cases. This includes design prototyping, marketing creatives, educational content, and product development. It is claimed to be able to synthesise information, structure visuals, and present them with clear layouts.

The company, however, notes that while Images 2.0 is a significant improvement, it is not without limitations. It can struggle with tasks requiring a highly accurate physical understanding, such as complex puzzles, origami instructions, or objects viewed from awkward angles.

It may also face challenges with very dense or repetitive visual details. Further, there could also be instances where outputs such as diagrams or labelled illustrations require manual verification for accuracy. Higher-resolution outputs beyond 2K are still in beta and may not always be consistent as well.