OpenAI Launches ChatGPT Images 2.0 and Codex Labs

OpenAI launched ChatGPT Images 2.0 on Tuesday, a significant upgrade to its image generation tool integrated into the ChatGPT chatbot, alongside a new enterprise training program called Codex Labs. The release, announced during a livestream event, aims to enhance the tool's accuracy and versatility for users worldwide.

The updated model, powered by the gptimage2 architecture, addresses longstanding limitations in AI image generation. It excels at rendering text within images, particularly in non-Latin scripts such as Japanese, Korean, Chinese, Hindi, and Bengali, where previous versions often produced distorted or inaccurate results. OpenAI officials stated that the tool now faithfully follows complex prompts, preserving details like small text, icons, user interface elements, and dense compositions. Users can generate images up to 2,000 pixels in resolution across flexible aspect ratios, from 3:1 wide to 1:3 tall, making it suitable for infographics, posters, and layouts.

A key advancement is the integration of reasoning capabilities, similar to those in ChatGPT's text processing. When users activate thinking or pro modes—available to paid subscribers—the model can search the...