Nano Banana 2 Is Here: Reasoning-Guided Image Generation, Now on Cencori

Google just released Nano Banana 2, and it's already live on Cencori. No waiting, no extra config — just swap the model ID and start generating.
Technically known as Gemini 3.1 Flash Image, Nano Banana 2 is the successor to Nano Banana Pro (Gemini 3 Pro Image). It brings a fundamentally different approach to image generation: reasoning-guided synthesis instead of traditional diffusion.
Why This Matters
Previous image models work by pattern matching — they denoise a random image until it looks like your prompt. Nano Banana 2 uses a multi-stage Plan → Evaluate → Improve loop, which means it actually reasons about what it's generating. The result: better logical consistency, fewer artifacts, and images that match your intent more closely.
What's New
Flash-Tier Speed, Pro-Level Quality
Nano Banana 2 combines the speed of Gemini Flash with the fidelity of Nano Banana Pro. You get photorealistic imagery with vibrant lighting, richer textures, and sharper details — without the latency tradeoff.
4K Native Resolution
Generate images natively from 512px all the way up to 4K. No upscaling hacks — the model generates at your target resolution from the start.
Production-Ready Text Rendering
Crisp, accurate text within images, including multi-language localization. If you've been frustrated by garbled text in AI images, this is a significant step forward.
Better Creative Control
Maintain subject consistency for up to 5 characters and fidelity for up to 14 objects in a single workflow. The model also pulls from Gemini's world knowledge and real-time web search for more contextually rich visuals.
Use It Today
Two model IDs are available through the Cencori gateway:
| Model ID | Description |
|---|---|
gemini-3.1-flash-image | Canonical ID |
nano-banana-2 | Alias (same model) |
SDK
import { Cencori } from 'cencori';
const cencori = new Cencori({ apiKey: process.env.CENCORI_API_KEY });
const result = await cencori.ai.generateImage({
model: 'nano-banana-2',
prompt: 'A futuristic city skyline at golden hour, neon reflections on glass towers',
size: '1024x1024'
});
console.log(result.images[0]);cURL
curl -X POST https://api.cencori.com/v1/images/generate \
-H "Authorization: Bearer $CENCORI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "nano-banana-2",
"prompt": "Product photography of a ceramic mug on a marble surface, soft natural light"
}'All Supported Image Models
For reference, here's the full lineup available through Cencori's image generation gateway:
| Model | Provider | Best For |
|---|---|---|
gpt-image-1.5 | OpenAI | Text rendering, highest ELO |
dall-e-3 | OpenAI | Creative, artistic images |
nano-banana-2 | Reasoning-guided, 4K, fast | |
gemini-3-pro-image | Photorealism | |
imagen-3 | High-quality generation |
Every request is automatically logged, rate-limited, and security-monitored through the Cencori gateway — whether you're using your own API keys (BYOK) or ours.
What's Next
We're watching for Nano Banana 2's GA release, at which point Google may stabilize the API model ID. If anything changes, we'll update the mapping transparently — your code stays the same.
Check out the Models Catalog for the full list of supported models, or dive into the Image Generation docs to get started.