OpenAI releases ChatGPT Images 2.0, a new image-generation model with improved text-rendering capabilities.

🔍 Let's dive in

OpenAI has released ChatGPT Images 2.0, touting major improvements over its predecessor. Sam Altman compared the generational leap to the jump from GPT-3 to GPT-5, though that claim comes directly from the company.

The model supports resolutions up to 3840×2160 pixels and introduces an outputQuality setting, priced at $30 per million output tokens. Its most concrete improvement is in rendering legible text within images — a longstanding weakness across AI image generators.

Independent testing found the model outperformed Google's Gemini and earlier OpenAI versions on complex compositional prompts, though it still struggles to reliably solve the visual puzzles it generates.

Synthesized across 2 sources

⚖️ Pros & cons

🕰 The timeline · 2 sources

Simon Willison analyst · 1d ago · 3/5

Where's the raccoon with the ham radio? (ChatGPT Images 2.0) ↗

OpenAI released ChatGPT Images 2.0 today, with Sam Altman claiming the improvement from the previous version is equivalent to the jump from GPT-3 to GPT-5. The new model supports higher resolutions up to 3840x2160 pixels and an outputQuality setting, with pricing at $30 per million output tokens. In testing with complex Where's Waldo-style image generation, the new model outperformed Google's Gemini and earlier versions, though it still struggles with reliably solving its own visual puzzles.

the leap from gpt-image-1 to gpt-image-2 was equivalent to jumping from GPT-3 to GPT-5

Looks like we definitely can't trust these models to usefully solve their own puzzles!

TechCrunch AI reporting · 1d ago · 3/5

ChatGPT’s new Images 2.0 model is surprisingly good at generating text ↗

OpenAI has released ChatGPT Images 2.0, its newest image-generation model. The model demonstrates significant improvements in generating images with readable text, a historically challenging task for image generators. The release reflects the rapid advancement of AI capabilities in visual content creation.

The diffusion models are reconstructing a given input. We can assume writings on an image are a very, very tiny part, so the image generator learns the patterns that cover more of these pixels.

Images 2.0 brings an unprecedented level of specificity and fidelity to image creation, able to follow instructions, preserve requested details, and render the fine-grained elements that often break image models: small text, iconography, UI elements, dense compositions, and subtle stylistic constraints.

🏷 Tags