ChatGPT Images 2.0
Article
TL;DR
OpenAI’s new image model sets a new bar for text rendering, photorealism, and mockup generation.
Key Takeaways
- High-quality 1536×1024 costs $0.165 — cheaper than gpt-image-1’s $0.25 at same size.
- First model practitioners find genuinely useful for slides, mockups, and UI prototypes.
- QR codes, chess boards, spatial grids still fail — spatial reasoning remains broken.
Discussion
Top comments:
-
[swalsh]: First image model actually useful for real work: slides and mockups
This is the first time i’ve found value in an image model for stuff I actually do. I’ve been using it to build powerpoint slides, and mockups. It’s CRAZY good at that.
- [ea016]: Detailed price comparison shows GPT Image 2 is cheaper per image than GPT Image 1
- [minimaxir]: Pokémon Pokédex grid prompt reveals domain knowledge + spatial layout limits
- [vunderba]: Gemini still leads on visual fidelity; OpenAI closed the gap on prompt adherence
- [simonw]: Quick Python snippet to test model via API with Where’s Waldo style prompt