Nano Banana 2 (Gemini 3.1 Flash Image)
Overview
Nano Banana 2 is Google DeepMind’s Flash-tier image model, announced 2026-02-26 as Gemini 3.1 Flash Image. It brings much of nano-banana-pro’s capability at Flash speed and became the default in the Gemini app, Flow, Search (AI Mode, Lens) and Google Ads. Closed model; API-only. Model ID gemini-3.1-flash-image (stable; it launched as gemini-3.1-flash-image-preview). Sibling: gemini-3.1-flash-lite-image (Nano Banana 2 Lite).
Specifications
- Resolutions 512px, 1K, 2K, 4K.
- Subject consistency: Google’s blog says up to 5 characters and 14 objects; API docs list up to 10 object + 4 character + 3 style references.
- Precision text rendering and in-image translation; stricter instruction following.
- Google Search grounding; controllable thinking (“minimal” or “high”).
- SynthID watermark on all outputs.
- Price: token-based on the Gemini pricing page (see https://ai.google.dev/gemini-api/docs/pricing). The old per-image figure, “4-6 s”, “#1 Artificial Analysis arena” and Elo figures are unverified and were dropped.
Access
Gemini app, Google AI Studio, Gemini API, Vertex AI, Flow (google-flow-nano-banana).
Prompting (practical)
Write full sentences, not keyword lists. Structure: subject, composition, action, location, style, editing instructions (what to preserve). Name the camera/lens and lighting, give exact text and placement, state aspect ratio and resolution, and refine over multiple turns.
Vs siblings
- Nano Banana (
gemini-2.5-flash-image): legacy, shutdown 2026-10-02. - Nano Banana Pro: 4K-capable too (the old table claiming Pro maxes at 2K was wrong); better for maximum fidelity.
Related
nano-banana, gpt-image-2-5, midjourney, flux-dot-1
Sources
- https://blog.google/innovation-and-ai/technology/ai/nano-banana-2/ (accessed 2026-09-30)
- https://ai.google.dev/gemini-api/docs/image-generation (accessed 2026-09-30)
- https://ai.google.dev/gemini-api/docs/models (accessed 2026-09-30)
- https://ai.google.dev/gemini-api/docs/pricing (accessed 2026-09-30)