ChatGPT Images 2.0 Beats Nano Banana 97-85
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- ChatGPT Images 2.0 scored 97% across nine image-generation tests, beating Google Gemini's Nano Banana, which scored 85% in a fresh head-to-head by ZDNET's David Gewirtz.
- OpenAI released ChatGPT Images 2.0 and GPT-5.5 last week; Images 2.0 goes beyond basic generation by adding the ability to include text and context derived from real data.
- In a December 2025 comparison, Nano Banana had scored 93% while ChatGPT managed just 74%, with ChatGPT's earlier poor showing attributed largely to its refusal to run pop-culture tests.
- Nano Banana stumbled on text rendering and prompt discipline, while Gemini's personalization capabilities raised privacy concerns in ZDNET's key takeaways.
- The nine tests covered photo restoration, black-and-white colorization, logo creation, studio portraits of a senior adult, a candid student portrait, a Back to the Future poster, and a Nightmare Before Christmas-style IT person.
- In the opening admiral photo recontextualization test, ChatGPT scored 14/15 and Nano Banana scored 12/15 after Gemini altered the subject's face (a wacky grin) and gave him a fuller beard.
Why it matters: ChatGPT's 97-to-85 reversal from its 74% score in December 2025 means OpenAI has closed and reversed the gap with Google's image-generation leader, with the win driven by tighter prompt discipline and stronger facial preservation; the remaining concern is that Gemini's personalization layer carries privacy trade-offs that ChatGPT doesn't currently offer.
