Google Nano Banana 2.1 Brings Cheaper AI Images To Gemini

Google has released Nano Banana 2.1, an update to its AI image generator aimed at producing more dependable pictures without the higher cost of its Pro model. The name is playful, but the practical pitch is serious: businesses and creators want better text in images, more consistent characters and products, and fewer failed attempts that waste time and money. Google’s model card, published on October 6, says the new system is built on Gemini 3.6 Flash.

In Google’s developer documentation, Nano Banana 2.1 is the recommended high-efficiency model for image generation and conversational editing. It supports 1K, 2K and 4K output, and Google says it improves visual quality, rendering of words inside pictures and consistency across successive edits. That last point matters when a marketer needs ten variations of the same campaign rather than ten attractive images that look like ten unrelated campaigns.

The model can draw on multiple reference images, helping it keep up to four characters and ten objects consistent in a workflow. It also supports grounding with Google Search. That could be useful for prompts that depend on current information, although a search connection should not be mistaken for a guarantee that every generated image accurately represents the world. Images made by the model carry Google’s SynthID watermark, according to the documentation.

The cost change is likely to be the more immediate selling point for developers. A comparison of Google’s posted API prices puts a 1K image at about 3.36 US cents with the new model, versus 6.7 cents for Nano Banana 2. A 4K image is listed at about 7.56 cents, down from 15.1 cents. Actual bills depend on the resolution, inputs and other API usage, but at large volumes those differences can change whether an image feature is worth offering at all.

For ordinary users, the distinction between the many Banana names is easy to lose. Nano Banana Pro is still Google’s premium choice for demanding creative work, while 2.1 is positioned as the faster, more economical everyday option. Our earlier look at Nano Banana 2 explained how Google began pushing high-quality image creation across its products. This release tries to make the follow-up cheaper and more reliable, not simply larger.

Google lists the new model for the Gemini app, AI Studio, the Gemini API, Search AI Mode, Ads, Flow and Stitch. Availability can vary by product and account, so users should check the tool they actually use rather than assume every feature arrives at once. The real test will be whether 2.1 reduces the small errors that force creators to regenerate an image, especially misspelled text, drifting faces and inconsistent products. A lower price helps; fewer do-overs would help even more.