When new AI models are announced, we usually see a slight ELO jump over previous SOTA models. That is not the case with ChatGPT Images 2.0: it is a massive leap over Nano Banana 2. According to Text to Image Arena, GPT-Image 2 has a rating of 1512, which is a big jump from Nano Banana 2’s 1271.
Exciting news – GPT-Image-2 by @OpenAI has claimed the #1 spot across all Image Arena leaderboards!
A clean sweep with a record-breaking +242 point lead in Text-to-Image – the largest gap we’ve seen to date.
– #1 Text-to-Image (1512), +242 over #2 (Nano-banana-2 with web-search… https://t.co/YYKjhgjhsn pic.twitter.com/IBN9a1RIJ4
— Arena.ai (@arena) April 21, 2026
The model is incredible with text. We were able to even get it to display a 15×15 grid of inventions over the years. While it is not perfect, it is by war closest to the prompt than any other model. This model has a huge lead over Nano Banana when it comes to single and multi-image edits:
I asked GPT Image 2.0 to draw me world changing inventions from 1000BC to 2026. It’s way too small for me to read but apparently they are there 🙂
Make a 15×15 grid starting with 1000 BC to 2026. In each section, show a unique invention diorama that changed the world at its… pic.twitter.com/sUtSLARbdw
— Gadgetify (@Gdgtify) April 21, 2026
What can you do with something like this? You can make amazing infographics, recipes, and even game assets. For example, you can use the image gen skill in Codex to speed up development.
GPT Image 2 + Codex is such a strong combo pic.twitter.com/d95R1rOAgV
— Angel 🌼 (@Angaisb_) April 21, 2026
This model now plans the composition, check its own work, and integrates instead of just one-shotting. That means you get better UI mockups, maps, and infographics.
