Nano Banana 2 vs. Nano Banana Pro: Differences and How to Use It

Nano Banana 2 is Google's new default AI image generator — officially Gemini 3.1 Flash Image — which replaced Nano Banana Pro as the standard model across the Gemini app, Google Search, AI Studio, the Gemini API, Google Cloud, Flow and Google Ads starting in late February 2026. The pitch is simple: Nano Banana Pro’s quality at Nano Banana (Flash)’s speed, with legible in-image text, built-in translation of text inside images, and consistency across up to five characters and 14 objects in a single generation.

If you've used Nano Banana Pro and are wondering what actually changed, or you've just seen "Nano Banana 2" pop up as the default option in Gemini and want to know how to use it, this explainer covers the differences, the pricing, where it's rolling out, and how to start generating with it.

What is Nano Banana 2?

Nano Banana 2 is Google's codename for Gemini 3.1 Flash Image, the image-generation and image-editing model that now sits at the center of Google's consumer and developer AI image products. In its launch announcement, Google describes it as combining Nano Banana Pro's "advanced knowledge and quality" with the speed of its Flash-tier models, so it renders at roughly Flash-model speed while aiming for output quality close to the higher-end Pro model.

It's the third model in the Nano Banana lineage: the original Nano Banana was Gemini 2.5 Flash Image, Nano Banana Pro followed as Gemini 3 Pro Image, and Nano Banana 2 is Gemini 3.1 Flash Image. Each generation has carried the "Nano Banana" nickname into mainstream use even though Google's actual product name is the Gemini model number.

Nano Banana 2 vs. Nano Banana Pro: what's different

The headline difference is positioning, not a simple upgrade: Nano Banana 2 is built to be Google's fast, default option, while Nano Banana Pro remains available as the higher-fidelity choice for people who need maximum factual accuracy or the largest native resolution. Google's own framing is that Nano Banana 2 narrows — but doesn't fully erase — the gap between Flash-level speed and Pro-level visual fidelity.

In practical terms: Nano Banana Pro still generates at the model Google calls its most capable for factual, detail-heavy work (infographics with dense data, diagrams that must be precise), and it remains reachable inside the Gemini app for Google AI Pro and Ultra subscribers through the three-dot regenerate menu. Nano Banana 2 is what everyone gets by default now, and Google says it carries over Pro's web-grounded knowledge, text rendering and instruction-following rather than being a stripped-down "lite" model.

Key features: text, translation, consistency and fidelity

Google highlights four capabilities as the core of the Nano Banana 2 pitch:

• Legible, accurate text rendering — for mockups, cards, posters and UI comps, where earlier image models routinely produced garbled or misspelled text.
• In-image translation and localization — Nano Banana 2 can take an image that already contains text (a sign, a label, a slide) and regenerate it with that text translated or localized into another language, while keeping the rest of the image intact.
• Character consistency — the model can keep up to five distinct characters looking the same across a single generation or an edit sequence, useful for storyboard-style or multi-panel outputs.
• Object fidelity — up to 14 distinct objects can be carried through a generation with their appearance preserved, which matters for product mockups and compositing tasks.

Google also says the model draws on Gemini's general knowledge and real-time web search grounding to render specific real-world subjects, places and data more accurately — the same capability that powers grounded infographics and diagrams in Nano Banana Pro. Google frames this as the difference between "a model that draws a generic bar chart" and one that can render a specific landmark, product or diagram correctly because it checked its own knowledge (or the live web) before generating pixels.

Google also says Nano Banana 2 follows complex, multi-part prompts more strictly than earlier models — a prompt that specifies lighting, camera angle, a specific character pose and a piece of on-image text all at once is more likely to come back with every instruction honored, rather than the model dropping one or two constraints. Output supports multiple aspect ratios and resolutions from 512 pixels up through 4K, depending on where you access the model.

Where Nano Banana 2 is available

As of its February 2026 rollout, Nano Banana 2 is live or in preview across:

• The Gemini app — default image model across Fast, Thinking and Pro modes, plus a new templates feature.
• Google Search — through AI Mode and Google Lens, on the Google app and in mobile/desktop browsers, with Google citing a rollout to 141 additional countries and territories and eight more languages at launch.
• Google AI Studio and the Gemini API — available in preview as gemini-3.1-flash-image-preview, documented in Google's developer post.
• Google Cloud — available in preview via the Gemini API in Vertex AI at launch, then moved to general availability through the Gemini Enterprise Agent Platform by late May 2026, alongside Nano Banana Pro.
• Flow — Google's AI filmmaking tool, where it's the default image model at zero credit cost to all users.
• Google Ads — powering image suggestions when advertisers build campaigns.

The general-availability milestone on Google Cloud also added a preview capability: Nano Banana 2 can now take a video file as an input prompt, alongside text, PDF and image references. Google calls this "deep video understanding" — the model analyzes visual context, specific subjects and actions in footage to produce context-aware images, thumbnails or infographics derived from the video. As of the Cloud announcement, this video-input feature and native 4K output were both still preview capabilities rather than fully generally available, while 1K and 2K output reached GA.

How much does Nano Banana 2 cost?

Nano Banana 2 isn't on the Gemini API's free tier; it's billed on the paid Standard tier like Nano Banana Pro, but at noticeably lower rates. Based on Google's published Gemini API pricing:

That puts Nano Banana 2 at roughly half the per-image cost of Nano Banana Pro at equivalent resolutions through the API. For most people, though, there's no separate "cost" to think about at all: generating images in the free Gemini app, Google Search's AI Mode, Lens or Flow doesn't require paying per image — the API pricing above applies to developers building on Nano Banana 2 directly through AI Studio, the Gemini API or Vertex AI.

Provenance: SynthID and C2PA Content Credentials

Every image Nano Banana 2 creates or edits carries an invisible SynthID watermark, Google DeepMind's technology for marking AI-generated media in a way that survives common edits like cropping, compression or filters without being visible to the human eye. Google says it has layered on C2PA Content Credentials as well, an industry-standard metadata format (backed by the Coalition for Content Provenance and Authenticity) that records whether and how AI was involved in creating a piece of media. Anyone can check whether an image was made with a Gemini model by uploading it in the Gemini app; Google says that SynthID-detection feature has been used more than 20 million times since it launched in November 2025, and C2PA verification inside the Gemini app was described as coming soon at the time of the Nano Banana 2 launch.

How to use Nano Banana 2

1. In the Gemini app: open a new chat, select the Fast, Thinking or Pro model, and describe the image you want. Nano Banana 2 generates by default; Google AI Pro and Ultra subscribers can still pull up Nano Banana Pro for a specific result via the three-dot menu and "regenerate."
2. In Google Search: use AI Mode or Google Lens from the Google app or a mobile/desktop browser — image generation and editing prompts route through Nano Banana 2 automatically where it's rolled out.
3. For developers: call the model as gemini-3.1-flash-image-preview through the Gemini API in Google AI Studio, or deploy it through Vertex AI / the Gemini Enterprise Agent Platform on Google Cloud for enterprise-grade SLA-backed usage.
4. In Flow: it's already the default image model for every user, at no credit cost, for generating reference frames and assets inside a video project.
5. With video input (preview): on Google Cloud, you can supply a video file as a prompt input alongside text to have the model extract context-aware stills, thumbnails or infographics from footage.

How Nano Banana 2 fits the wider AI-image field

Nano Banana Pro's debut already reshaped how outlets benchmark AI image tools against rivals like GPT Image 2 and Grok Imagine on text rendering, character consistency and factual grounding. Nano Banana 2's job is to bring that same bar down to the speed tier most people actually use day to day, rather than reserving it for a premium subscription model. Google has not published independent benchmark scores comparing Nano Banana 2 directly against third-party models, so any such comparison should be treated as a separate, ongoing storyline rather than a settled result.

What's next

Google has continued iterating on the Nano Banana family since the Nano Banana 2 launch, including a lower-cost "Lite" variant aimed at high-volume, cost-sensitive use cases, and the broader push to general availability for both Nano Banana 2 and Nano Banana Pro on the Gemini Enterprise Agent Platform. Video-file input and native 4K output remain preview-only capabilities for now, which suggests Google is still tuning reliability before promoting them to general availability.

For most people the practical takeaway is simple: you don't need to pick a model at all. Nano Banana 2 is already what loads by default in the Gemini app, AI Mode and Lens, so the "upgrade" happens automatically. The decision point only really shows up for developers choosing between the Flash-tier and Pro-tier model in the API, or for Google AI Pro/Ultra subscribers deciding whether a specific image is worth manually regenerating on Nano Banana Pro. Expect further updates to rollout countries, pricing tiers and the video-understanding feature as Google moves those capabilities out of preview and toward general availability.

Written byTheo Park

Theo Park runs the AI desk at Pandromeda. He follows model launches from the frontier labs and the open-weight community, tracks the assistants and developer tools built on them, and explains what each release changes on pricing, capability and safety. His reporting leans on primary sources: model cards, technical reports, API documentation and the companies' own announcements.

More from Theo ParkAll AI coverage