Blind test of 200 real image edits: nano banana pro loses to its cheaper base model
ravisparikh · x · 2026-10-17
A team ran 200 real image edits through 8 image models, judged blind by Claude, Gemini and GPT: nano banana pro ($0.15) scored worse than regular nano banana ($0.045); flux.3 klein ($0.009) is 17x cheaper than pro and within 0.14 points of it; krea 3 edit, priced like nano banana, lost head-to-head 157-208; the real winner was gpt image 2.5 flare with 71% of blind matchups. Takeaway: stop paying for the logo.
More from Models
- Google AI Studio rolls out Gemini 4 Argon for Pro subscribers — jocarrasqueira · 2026-10-17
- Runway's internal study: Solaris beats Claude Opus 5 on instruction-following 62% vs 25% — _AustinCalvert_ · 2026-10-17
- GPT-6 Astra nears 100% on Columbia's SRE-Bench reverse-engineering benchmark, others below 60% — banteg · 2026-10-17
- Musk says connecting Grok to your X DMs is the best way to use it — elonmusk · 2026-10-17
- Alexandr Wang: Meta MSL built 'a machine that builds models' in one year — alexandr_wang · 2026-10-17
- LLMs pick 7 in 30/30 tries: EMNLP paper traces the bias to SFT training data — radamihalcea · 2026-10-17