Google's Nano Banana 2 Lite cuts AI image costs and speeds up output

Google has introduced Nano Banana 2 Lite, a cheaper and faster AI image model aimed at users who care about cost per image as much as output quality. For retail investors watching the AI buildout, the launch is another sign that major platforms are pushing generative AI from premium tools into high-volume, lower-cost workflows.

The model’s official name is gemini-3.1-flash-lite-image. According to Google, it sits at the bottom of the company’s current image generation lineup, below Nano Banana 2 and Nano Banana Pro. It also replaces the original Nano Banana model, gemini-2.5-flash-image.

Nano Banana 2 Lite produces text-to-image results in about four seconds, according to Decrypt, making it 2.7 times faster than Nano Banana 2. Text-to-image means a user writes a prompt, and the model creates an image from that written instruction.

The economics are the headline for anyone thinking about scale. Decrypt reported that Nano Banana 2 Lite costs about $0.034 per image at 1K resolution, compared with roughly $0.067 for Nano Banana 2 at the same resolution. In plain terms, a user running the Lite model would spend about half as much per image before considering any other platform or usage costs.

Where the model fits

Google is distributing Nano Banana 2 Lite through Google AI Studio, the Gemini API and the Enterprise Agent Platform, according to Decrypt. It is also being built into consumer products including Search, the Gemini app, NotebookLM and Google Photos.

The model can also work with Gemini Omni Flash, Google’s video generation model, through the Interactions API. An API, or application programming interface, lets software systems talk to each other. In this case, Decrypt reported that the Interactions API allows users to apply as many as three sequential edits in one session.

The lineup now has a clearer structure: Lite is positioned around speed and cost, Nano Banana 2 is framed as the middle option for quality and speed, and Nano Banana Pro is aimed at more complex professional work, according to Decrypt.

How pricing compares

Nano Banana 2 Lite is priced close to Seedream 5.0 Lite, which Decrypt said costs about $0.031 to $0.035 per image. Reve 2.0 is cheaper at roughly $0.0067 per image through its API, though Decrypt noted that it does not offer the same deployment reach as Google’s infrastructure. Qwen Image Edit was described by Decrypt as a free, open-source option for standard uses.

For businesses or creators producing many images, small price differences can add up quickly. A lower per-image cost matters most when a workflow involves repeated drafts, bulk creative testing or automated image generation inside a product.

Quality is the trade-off

Decrypt tested Nano Banana 2 Lite and Nano Banana 2 with the same prompts across five categories. The testing found that the Lite model matched or beat Nano Banana 2 in many areas, but that the higher-priced model may still be stronger when fine detail matters.

In one realism test, both models were asked to create a cinematic rooftop portrait of a 32-year-old female architect at sunset, with specific instructions for clothing, glasses, blueprints held in her left hand, city background, lighting, lens effect, skin texture and film grain. Decrypt said Nano Banana 2 Lite met the basic prompt requirements, including the outfit, rooftop setting, glasses, blueprints and blurred skyline.

The weaker points showed up under closer review. Decrypt reported that the Lite output gave the subject only one hand, made that hand oversized, rendered the rim light faintly and produced skin texture that looked acceptable at thumbnail size but weaker up close. The result was described as closer to a competent stock-style image than a cinematic portrait.

That split explains the product strategy: Google is lowering the cost and wait time for everyday image generation, while keeping more detailed work tied to its higher-tier models.

This story draws on original reporting from Decrypt.