What Is Nano Banana 2? Gemini 3.1 Flash Image Explained (April 2026)

Google DeepMind's Nano Banana 2, officially named Gemini 3.1 Flash Image, landed on February 26, 2026 and quietly reset the price-performance curve for image generation. Google's pitch for Nano Banana 2 is short: take the capabilities from Nano Banana Pro (world knowledge, precise text rendering, multi-character consistency), and run them on the Flash inference stack at roughly half the cost.

This guide explains what Nano Banana 2 actually is, what Google has officially confirmed about Nano Banana 2, and when it makes sense to pick Nano Banana 2 over the regular Nano Banana or the higher-tier Nano Banana Pro. Every Nano Banana 2 spec below is sourced to a Google blog post, DeepMind product page, or ai.google.dev documentation page.

TL;DR

• Official name: Gemini 3.1 Flash Image. Codename: Nano Banana 2.
• Released: February 26, 2026 (Google Workspace Updates).
• API model ID: gemini-3.1-flash-image-preview (ai.google.dev).
• Positioning: Pro-tier capabilities (world knowledge, text rendering, character consistency) at Flash-tier speed and price.
• Pricing: $60 per 1M output tokens, roughly $0.067 per 1K image, $0.101 per 2K, $0.151 per 4K (Gemini API pricing).
• Where it runs: Gemini app, Google Search (AI Mode and Lens), AI Studio, Gemini API, Vertex AI, Flow, Google Ads.

What is Nano Banana 2?

Nano Banana 2 is the cover name Google uses publicly for its newest image generation and editing model, Gemini 3.1 Flash Image. The Google blog announcement frames Nano Banana 2 as combining "the advanced world knowledge, quality and reasoning you love in Nano Banana Pro, at lightning-fast speed" (blog.google).

The Nano Banana 2 naming convention follows the same pattern Google used for Nano Banana Pro (which is gemini-3-pro-image-preview under the hood): a friendly product nickname for consumer surfaces, and a formal gemini-3.x-...-image identifier for the API. So when you see "Nano Banana 2" in the Gemini app and gemini-3.1-flash-image-preview in your API code, they are the same model.

Nano Banana 2 replaces the earlier Nano Banana (the original Gemini 2.5 Flash Image release) in the Gemini app across the Fast, Thinking, and Pro tiers, per Google's own announcement.

Release date and rollout

Free Gemini app users get image generation up to 1K resolution; paid Workspace and Gemini subscribers get up to 2K, per the Workspace Updates post. Image generation is available to users under 18; image editing is restricted to 18+.

Key features Google has confirmed

Every claim in this section comes directly from Google's blog or DeepMind's product page. Anything I could not source is omitted.

1. Precision text rendering and translation

DeepMind documents that Nano Banana 2 can "render legible text directly into your images" with control over fonts, styles, and sizes, pitched at greeting cards, marketing mockups, and posters (deepmind.google). Nano Banana 2 also supports in-image translation and cultural localization, so the same poster can be regenerated with text in another language without leaving the model.

2. World knowledge and web grounding

Nano Banana 2 is "powered by real-time information and images from web search to more accurately render specific subjects and create infographics and data visualizations" (blog.google). In practice this is the same architectural feature that lets Nano Banana Pro generate up-to-date charts and reference real subjects, now available at Flash latency inside Nano Banana 2.

3. Subject consistency: 5 characters, 14 objects

DeepMind's product page states Nano Banana 2 maintains "character resemblance of up to five characters and the fidelity of up to 14 objects" within a single workflow. That is the identical figure published for Nano Banana Pro. DeepMind explicitly carried the consistency capability across to the Flash tier rather than reserving it for Pro.

4. Resolutions: 512px through 4K

The supported Nano Banana 2 output sizes documented in the image generation API reference are:

• 512 (a new lower tier specific to Gemini 3.1 Flash Image)
• 1K (default)
• 2K
• 4K

The 512 option is the notable Nano Banana 2 addition. It makes high-volume thumbnail or icon workflows cheaper than anything Pro offers. Aspect ratios supported are: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9 (14 ratios total).

5. Configurable thinking levels

Per the API reference, Nano Banana 2 exposes two thinking_level values: minimal (default, lowest latency) and high. Thinking tokens are billed regardless of whether you ask the API to surface the thoughts in the response. This is a Flash-tier handle on the same reasoning machinery that powers the Pro model.

6. SynthID watermarking and C2PA Content Credentials

Google reports that every image generated by Nano Banana 2 carries an invisible SynthID watermark for AI-generated content identification, paired with C2PA Content Credentials metadata (blog.google). This matches the provenance setup used across the Gemini image lineup, and Nano Banana 2 ships it on by default.

Pricing: Nano Banana 2 is about half the cost of Pro

Nano Banana 2 pricing comes straight from the Gemini API docs. Nano Banana 2 charges per output token; image sizes consume a fixed token count, so the per-image rate is deterministic.

Input pricing: $0.50 per 1M tokens for text and image input (Standard), $0.25 per 1M for Batch.

For comparison, Gemini 3 Pro Image (Nano Banana Pro) is listed on the same pricing page at $0.134 per 1K/2K image and $0.24 per 4K image. So Nano Banana 2 at 1K is roughly half the price of Nano Banana Pro at 1K, and Nano Banana 2 at 4K is roughly 60% the price of Pro at 4K.

Nano Banana 2 vs Nano Banana vs Nano Banana Pro

Three Google image models are now in active deployment, which is one too many before you sort out which is which:

Sources: Gemini 3.1 Flash Image page, Gemini Image / Nano Banana page, Nano Banana Pro page, and the Gemini API pricing page.

When to pick which

• Pick the original Nano Banana if you have a working integration on gemini-2.5-flash-image and have not yet had time to migrate to Nano Banana 2. Otherwise there is no use case where it beats Nano Banana 2; Nano Banana 2 is strictly a superset at lower or equal price.
• Pick Nano Banana 2 for almost everything else: marketing creative, multilingual posters, product mockups, infographics, character-driven series, anything at 1K or 2K. The 512px tier also makes Nano Banana 2 the cheapest Google option for high-volume thumbnail and icon work.
• Pick Nano Banana Pro for the highest-stakes deliverables where the extra reasoning time genuinely buys you something Nano Banana 2 cannot. DeepMind keeps Pro available "for specialized tasks" via regeneration on Pro and Ultra plans. Concrete cases: dense infographics that combine many entities, hero brand imagery where the brief is unforgiving, or workflows already locked to gemini-3-pro-image-preview.

For a deeper Pro-vs-OpenAI comparison, see our Nano Banana Pro vs GPT Image breakdown.

Where Nano Banana 2 runs

Google has confirmed Nano Banana 2 availability across these surfaces (blog.google):

• Gemini app: replacing Nano Banana Pro across the Fast, Thinking, and Pro models.
• Google Search: both AI Mode and Lens.
• Google AI Studio and the Gemini API: preview, paid API key required (build with Nano Banana 2).
• Google Cloud Vertex AI: preview, for enterprise deployments.
• Flow: zero-credit access for all users.
• Google Ads: built into the asset generation flow.

The fastest way to test Nano Banana 2 without setting up a Google Cloud project is the Gemini app (free tier, 1K resolution) or AI Studio (paid API key, all resolutions). For a single dashboard that runs Nano Banana 2 alongside other image models, gptimg.co/nano-banana-2 wraps the Nano Banana 2 API directly.

Limitations Google acknowledges

DeepMind's product page lists known weaknesses of Nano Banana 2 plainly, worth knowing before you ship:

• Small face rendering can be imperfect.
• Spelling inaccuracies are still possible despite the text-rendering upgrade.
• Factual accuracy on data-driven content has limits.
• Complex edits and dramatic lighting changes can produce inconsistent results.

These are honest Nano Banana 2 qualifications from Google itself, not third-party complaints. For mission-critical text inside an image, generate, read it back, and regenerate if any character is wrong. Nano Banana 2 is much better than DALL·E 3 ever was at in-image text, but not yet typo-proof.

Frequently asked questions

What is the official name of Nano Banana 2?

The official model name is Gemini 3.1 Flash Image. "Nano Banana 2" is the consumer-facing nickname Google uses in the Gemini app, on the DeepMind product page, and in marketing. The API identifier is gemini-3.1-flash-image-preview (ai.google.dev).

When did Nano Banana 2 launch?

Nano Banana 2 launched on February 26, 2026, simultaneously across the Gemini app, Google Search, AI Studio, Vertex AI, Flow, and Google Ads (Workspace Updates).

How much does Nano Banana 2 cost?

On the standard Gemini API, Nano Banana 2 output is billed at $60 per 1M tokens, which works out to $0.045 per 0.5K image, $0.067 per 1K, $0.101 per 2K, and $0.151 per 4K. The Batch API offers a flat 50% discount on Nano Banana 2 charges. Source: Gemini API pricing.

Is Nano Banana 2 better than Nano Banana Pro?

Not strictly. Pro retains the deeper reasoning tier and remains available for specialized tasks. What Nano Banana 2 does is bring the Pro-tier feature set (world knowledge, text rendering, 5-character consistency, 4K output) to a Flash-tier price (~half the per-image cost) and Flash-tier latency. For most production workloads at 1K or 2K, Nano Banana 2 is the more economical pick; Pro is reserved for the very highest-stakes outputs that Nano Banana 2 cannot quite carry.

What resolutions does Nano Banana 2 support?

512 (0.5K), 1024 (1K, default), 2048 (2K), and 4096 (4K). It also supports 14 aspect ratios including 1:1, 9:16, 16:9, 21:9, 4:1, 1:4, 8:1, and 1:8 (API reference).

Where can I try Nano Banana 2 free?

Inside the Gemini app on a free Google account (capped at 1K resolution), or in Google AI Studio with a paid API key. The studio at gptimg.co/nano-banana-2 also runs the model with trial credits on signup.

Does Nano Banana 2 watermark its images?

Yes. Every Nano Banana 2 image carries an invisible SynthID watermark identifying it as AI-generated, paired with C2PA Content Credentials metadata for provenance (blog.google).

Try Nano Banana 2

Try Nano Banana 2 free at /nano-banana-2. The model runs directly in your browser with trial credits on signup. No Google Cloud project, no API key setup. For a side-by-side against the Pro tier, head to /nano-banana-pro; to compare with the original 2.5 Flash release, see /nano-banana.

Sources

• Nano Banana 2: Combining Pro capabilities with lightning-fast speed, Google blog announcement
• Build with Nano Banana 2, our best image generation and editing model, Google developer blog
• Gemini 3.1 Flash Image – Nano Banana 2, DeepMind product page
• Introducing Nano Banana 2 in the Gemini app, Google Workspace Updates
• Image generation with the Gemini API, ai.google.dev API reference
• Gemini Developer API pricing, ai.google.dev pricing page
• Nano Banana Pro / Gemini 3 Pro Image, DeepMind product page
• Gemini Image — Nano Banana, DeepMind original Nano Banana page

Last reviewed against source pages: 2026-04-18. Pricing and capability figures change periodically; confirm in the linked sources before acting on the numbers above.