Google put Nano Banana 2.1 into general availability on 6 October 2026. It is rolling out in the Gemini app, Search AI Mode, Google Ads, and developer tools including the Gemini API, AI Studio, Flow, and Stitch. The API model id is gemini-nano-banana-2.1. It replaces Nano Banana 2 (gemini-3.1-flash-image), which Google has deprecated.

What actually happened

The price cut is on image output, and it is not uniform. On Google’s Gemini API pricing page, standard image output for 2.1 is $30 per 1 million tokens, down from $60 on Nano Banana 2. A 1K image (1024×1024, 1,120 tokens) is $0.0336, against $0.067 before. A 2K image is $0.0504, against about $0.101. A 4K image consumes 3,780 tokens and lands at about $0.113, against $0.151. That is roughly a quarter off, not half. Decrypt’s $0.0756 figure for 4K does not match the token count on the pricing page.

Batch is another 50% off image output, at $15 per 1 million tokens. A 1K batch image is $0.0168. Input is not cheaper. Text, image, and video input is $1.50 per 1 million tokens, up from $0.50. Text and thinking output is $7.50, up from $3. Google also says 2.1 handles up to 14 reference images, keeps up to four characters and ten objects consistent, and can ground a generation in Google Search. Those quality claims are Google’s own tests. Independent scores were not out at launch.

Why a small team cares

A 1K or 2K ad batch is the only place the cut shows up this week.

Take a Shopify brand that already generates catalog and ad variants in Gemini or through the API. Previously, 1,000 standard 1K images were about $67 in image output. Now they are $33.60, or $16.80 on Batch. A 2K social set falls the same way. A 4K pack does not. Prompt-heavy jobs with several reference images give some of that back, because input is triple the old rate. Google Ads users do not see a per-image invoice inside the ad account. The cut lands if the creative is produced through the API or a paid Gemini plan that meters image generation. App and Search use still sits inside plan limits.

Hype vs useful

“Half price” is true for 1K and 2K image output on the standard paid tier. It is not true for 4K, and it is not the whole call. Thinking tokens and reference images can eat the gap on a short edit loop. Nano Banana 2 Lite was already $0.0336 at 1K, so a shop that only needed cheap 1K squares may see no image-output change at all. Quality is the open item. Google’s preference score is not a substitute for your product-on-white and text-in-image rejects. Run one batch. Keep the old model on the jobs 2.1 still fails.

Source: decrypt.co (https://decrypt.co/380257/google-launches-nano-banana-2-1)