Nano Banana 2 Lite: Google launches faster, cheaper image model

Google has released a new AI image generation model aimed at developers and enterprise teams who need speed and low costs above all else. The model is officially called Gemini 3.1 Flash-Lite Image but is marketed under the name Nano Banana 2 Lite. It generates images at 1,000 pixel resolution in around four seconds and is priced at $0.034 per 1,000 images. Google announced the release in a post on The Keyword blog and the model is available immediately through Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform (GEAP).

Alongside its developer release, Google is rolling out Nano Banana 2 Lite across several consumer products, including AI Mode in Search, the Gemini app, NotebookLM, Google Photos, and Google Ads. All images generated by the model carry a SynthID watermark, an invisible digital identifier that marks content as AI-generated.

The model is positioned as the entry-level option within Google’s Nano Banana image model family. It replaces Nano Banana (Gemini 2.5 Flash Image) and undercuts it on price: the previous model cost $0.039 per 1,000 images. The standard Nano Banana 2 costs $0.067, and the top-tier Nano Banana Pro sits at $0.134. According to Google’s internal notes cited in the sources, Nano Banana 2 Lite delivers roughly 60 to 70 percent of the capability of the higher-tier models while running significantly faster.

What the model can and cannot do

Nano Banana 2 Lite supports only 1,000 pixel resolution output, whereas the standard Nano Banana 2 and Nano Banana Pro both offer 1,000, 2,000, and 4,000 pixel options. Despite this limitation, the model scores competitively in independent benchmarks. On the Text to Image arena Elo scale, it achieves 1,251, ahead of the older Nano Banana (1,151) and even the more expensive Nano Banana Pro (1,245). For image editing tasks, it scores 1,308 for single-image edits and 1,294 for multi-image edits.

Google highlights three capabilities it considers particularly strong in this model:

• World knowledge for generating contextually accurate scenes and location-specific mockups
• Character consistency across sequential image generations, useful for storyboarding or product visualisation
• Text rendering for embedding legible copy directly into generated images, including across multiple languages

Google also notes that editing existing images may take slightly longer than generating new ones from scratch, due to the additional processing required to modify existing pixels.

Known limitations include difficulties with small faces, accurate spelling, and fine details. When generating infographics or data visualisations, the model may produce factually incorrect results. Google recommends always verifying such outputs manually.

Who it is built for

Google describes Nano Banana 2 Lite as a tool for high-throughput commercial workflows rather than artistic or high-end creative production. Typical use cases include automated advertising asset generation, real-time A/B testing of visual content, rapid layout prototyping, and dynamic content in digital commerce applications.

The model runs exclusively through Google’s managed cloud infrastructure. Developers cannot download or host it independently, which keeps operational complexity low but ties usage to Google’s pricing terms. As Carl Franzen reports for VentureBeat, this approach reflects a deliberate strategy to lock enterprise developers into Google’s commercial platform ecosystem.

Sources

• Google unveils Nano Banana 2 Lite aka Gemini 3.1 Flash-Lite for low cost, 4-second fast enterprise image generations – VentureBeat
• Nano Banana 2 Lite – Google DeepMind
• Start building with Nano Banana 2 Lite and Gemini Omni Flash – The Keyword, Google Blog