OpenAI has officially launched ChatGPT Images 2.0, featuring substantial enhancements in AI image generation capabilities. This release follows extensive testing under the name “duct tape” on LM Arena AI, allowing users to create complex images that include long text blocks, user interfaces, and even floor plans.
The new model, which incorporates the gpt-image-2 architecture and “Thinking” features for ChatGPT subscribers, marks a significant shift in OpenAI’s approach to visual media. OpenAI emphasized, “Images are a language, not decoration,” highlighting a move toward viewing images as functional tools rather than mere visual elements.
While specific benchmarks were not disclosed, OpenAI claims ChatGPT Images 2.0 operates at “state-of-the-art” levels. This release comes amid increasing competition from Google’s Nano Banana 2 model, which debuted in February 2026. Early testing suggests that ChatGPT Images 2.0 can generate user interfaces and image packs with higher fidelity than its competitors.
OpenAI is deprecating GPT-Image-1.5 as the default model but will keep it available via the API. This indicates confidence in the new model’s capabilities for both casual and high-value tasks. ChatGPT Images 2.0 now integrates OpenAI’s “O-series” reasoning, allowing the model to research and plan before generating images.
During demonstrations, the model effectively produced high-quality educational materials, including maps of ancient empires with detailed legends. This functionality aims to support users in creating informative and visually accurate content.
Improvements in typography enable the model to produce legible text even in intricate designs. Additionally, ChatGPT Images 2.0 supports generating images in multiple languages, addressing a historical bias in AI-generated imagery.
Users can now create up to eight coherent images from a single prompt, maintaining consistency and continuity among the generated outputs. The rollout includes a tier-based access structure: free users can utilize base features, while Plus and Pro subscribers gain access to advanced capabilities, including web search and multi-image generation. API pricing has decreased, with costs set at $8.00 for image generation outputs.
OpenAI has implemented a “multi-layered stack” of safety protocols to ensure the responsible use of AI-generated images, including watermarking and content filtering. In response to inquiries regarding potential misuse, OpenAI stated it takes safety measures regarding political influence and advertising campaigns seriously.
As ChatGPT Images 2.0 transitions into active use, it aims to redefine the interaction between visual generation and user intent through enhanced reasoning and contextual understanding. OpenAI’s objective is to assist users in executing economically valuable creative tasks while balancing speed and quality.
