Google's new image model Nano Banana 2.1 generates better images for less money
Google's Nano Banana 2.1 beats Pro on benchmarks but not in the real world—a reminder that metrics and practice diverge.
Why it matters
Google ships a smaller, cheaper image model that scores higher on benchmarks than its predecessor Pro, but the article itself notes that Pro often produces better images in practice—highlighting the persistent gap between benchmark wins and user-perceived quality in generative AI.
The key facts
11 to knowNano Banana 2.1 uses Gemini 3.6 Flash backend
Beats previous Pro model in some benchmarks
Lower cost than Pro
Pro often produced better images in practice despite lower benchmark scores
Predecessor Nano Banana also scored well in tests
Model: Google Nano Banana 2.1
Powered by Gemini 3.6 Flash
Outperforms previous Pro model on some benchmarks
Costs less than Pro model
Predecessor often produced better images in practice despite lower benchmark scores
Benchmark-to-practice quality divergence documented
The story so far
Earlier coverage of this storyline
- Nano Banana 2.1 now available on AI GatewayVercel Blog
- This story
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Google's new Nano Banana 2.1 image model uses Gemini 3.6 Flash and beats the previous Pro model in some benchmarks at a lower cost. But its predecessor also scored well in tests, while Pro often produced better images in practice.