Google launches Nano Banana 2.1, cutting image prices

By VK

Published 7 Oct 2026, 11:27 UTC

Reporting from The Decoder

Google's Nano Banana 2.1 image model costs $0.0336 per 1K image in the Gemini API, half the Nano Banana 2 price, and is rolling out across Google apps.

What it means for founders

Rerun your unit economics. If your product generates high volumes of 1K or 2K images, output cost per image halves. But the higher input and thinking prices eat into that for edit-heavy flows that send several reference images per call, so measure cost per finished image, not the headline rate.
Test the cheap path first. Compare 2.1 at minimal thinking against medium on your own prompts. If minimal holds quality, you get the lowest bill; if not, you know what quality costs.
Use Batch for anything not live. Catalog shots, ad variants and thumbnails can run through the Batch API at $0.0168 per 1K image.
Plan the migration now. Google is steering Nano Banana 2 users to 2.1, so watch its deprecations page for a shutdown date on the old model ID. Pin model IDs, run side-by-side tests on your hardest cases, especially small text and multiple characters, and keep Pro as a fallback where realism decides the sale.

The story

Google released Nano Banana 2.1 on October 6, 2026, an image generation and editing model that replaces Nano Banana 2 and cuts the per-image price of a standard 1K picture in the Gemini API by half. The model, listed as gemini-nano-banana-2.1 on Google's Gemini API model page, is built on Gemini 3.6 Flash and is reaching Google AI Studio, Flow, Stitch, Google Ads, Search's AI Mode and the consumer Gemini app.

Nano Banana 2.1 pricing in the Gemini API

Google now charges $30 per million image output tokens, down from $60 for Nano Banana 2. On its pricing page that works out to:

1K image: $0.0336, versus $0.067 before
2K image: $0.0504, versus $0.101
4K image: $0.113, versus $0.151, a smaller cut because a 4K image now uses 3,780 tokens instead of 2,520
Batch API: half the standard rate, so $0.0168 per 1K image

Other line items move the other way. Input costs $1.50 per million tokens instead of $0.50, and text and thinking output rises from $3 to $7.50. There is no free tier and no 0.5K size on the price list. Nano Banana Pro still costs $0.134 per 1K or 2K image.

What Google says changed

Google claims better realism, prompt adherence, text rendering and character consistency across turns. The model accepts up to 14 reference images, holds as many as 4 characters and 10 objects steady across outputs, and can pull from Google Web and Image Search. Developers pick a thinking level: minimal, medium (the default) or high. Google also says it fixed tiling artifacts on very wide panoramic formats.

In Google's own human preference ratings, the thinking version scores 1050 against 990 for Nano Banana 2 and 935 for Nano Banana Pro. Independent testing is less flattering: The Decoder found Pro still often produced more natural images, with 2.1 getting scale wrong in its benchmark prompt. Google's model card lists blurry small text at 1K and left and right confusion among known limits.

What we don't know yet

When the old gemini-3.1-flash-image endpoint will shut down. Google names 2.1 as its replacement, but its deprecations page lists no shutdown date.
Whether outside, blind comparisons will back up Google's preference scores.
How much thinking tokens add to the real cost of each image at the medium default.