Nano Banana 2 Review: What's New and Is It Worth Switching?

Another day, another model drop.

Google just released Nano Banana 2, which is powered by their new Gemini 3.1 model. It is faster, cheaper, and better than the previous Nano Banana model. Quick overview below plus a bunch of example prompts I’ve run comparing the new model to the old one (logos, infographics, images, etc.)

What’s new with. Nano Banana 2

Google’s launch post is packed with feature bullets. Here’s the practical translation.

Advanced world knowledge / grounding

The model can pull in real-time information about the world. This means you could have it generate an image of New York every day and it would match the weather for that day. GPT-image does this as well, but still cool.

2) Precision text rendering + translation
Image models have historically been awful at text. Nano Banana 2 is explicitly pushing:
- legible text for mockups, signs, ads, cards
- translation + localization inside images

If this actually holds up, it unlocks a ton of knowledge work use cases. More examples later on.

3) Creative control upgrades
The big one here is consistency and instruction following. The feature that jumped off the screen for me here was subject consistency. Keeping characters consistent across generations is a huge upgrade and makes it much easier to string together things like storyboards.

My tests

I tested in Converge using our Nano Banana component which makes it super easy to drop image generation via Nano Banana into any app.

I built a quick image studio app so I could compare the three latest Nano Banana models:
- Nano Banana 2 (new): gemini-3.1-flash-image-preview
- Nano Banana Pro (older “best”): gemini-3-pro-image-preview
- Older Flash: gemini-2.5-flash-image

Corgis skateboarding in SF

Nano Banana 2 looks the crispist by far.

Uploaded the Converge logo and made it fuzzy

Nano Banana 2 and Nano Banana Pro pretty close here in quality and speed.

Prompt: Transform a simple flat vector logo into a soft, 3D fluffy object. Use the exact colors. The shape is fully covered in fur, with hyperrealistic hair texture and soft shadows. The object is centered on a clean, light gray background and floats gently in space. The style is surreal, tactile, and modern, evoking a sense of comfort and playfulness. Studio lighting, high-resolution render.

Infographic / diagram generation

I took the water cycle image from Google’s blog post and had Nano Banana(s) recreate it for photosynthesis.

This is the clearest win for Nano Banana 2.

Creating Meta Ads

For this one I took a screenshot of the Converge homepage and told Nano Banana to make some Meta Ads. For such a lazy prompt I was really impressed. Tbh, I kinda like Nano Banana Pro more than Nano Banana 2.

Speed and Cost

Nano Banana 2 is about 50% cheaper than its predecessor and is supposed to be much faster, although I haven’t seen much of a speed difference.

Where it’s available

Google is rolling Nano Banana 2 out broadly across:

• Gemini app
• Search (AI Mode, Lens)
• Gemini API
• Vertex AI (preview)
• Flow (default)
• Google Ads (creative suggestions)

Nano Banana 2 is also live to test in Converge! Just use the Nano Banana component and you’ll be good to go.