Google DeepMind

Nano Banana 2

Google’s fast, photoreal-leaning image model. Cheap per gen, great for iteration.

Browse Nano Banana 2 prompts →Compare side-by-side

3 benchmarks runLatest: google/nano-banana-2

Nano Banana 2 (the codename for Gemini’s Flash image model) is the fast counterpart to GPT Image 2. Typical generations land in 15–30 seconds at a lower per-gen cost, which makes it the right pick when you’re iterating on style or generating in volume. The trade-off is precision: it interprets prompts loosely and is less reliable for exact-text or pixel-perfect compositional work.

Strengths

✓Fast: 15–30 seconds per generation in typical conditions.
✓Strong photoreal style and character consistency.
✓Lower per-generation cost than GPT Image 2.

Weaknesses

✗Text rendering is inconsistent — don’t rely on it for signage.
✗Less literal: invents details you didn’t ask for.
✗Higher run-to-run variance than GPT Image 2.

Nano Banana 2 across our benchmarks

AI Movie Poster

See all models on this prompt →

Vintage Bookshop Neon Sign

See all models on this prompt →

Dramatic Mountain Silhouette

See all models on this prompt →

Want to use Nano Banana 2 on your own photos?

We host Nano Banana 2 alongside every other major AI image model. Free to try.

Open the Nano Banana 2 prompt library →