# Google Ships Nano Banana 2.1, Halving 1K Image Costs

Published 2026-10-07 · AI.info Team · AI.info

> Google’s Nano Banana 2.1 became generally available on October 6, with standard 1K image output priced at $0.0336, half the predecessor’s $0.067 charge. The update improves image quality and text rendering and fixes tiling artifacts in wide and panoramic formats, while search grounding was already listed for Nano Banana 2.

## Lower prices, but not by the same amount

[Google](https://ai.google.dev/gemini-api/docs/models/gemini-nano-banana-2.1) released Nano Banana 2.1 on October 6, pitching a more capable image model at lower API image-output prices than its predecessor. The cut is sharpest at smaller sizes: a standard 1K image costs $0.0336, down from $0.067 for Nano Banana 2. At 4K, the new image-output charge is $0.113, compared with $0.151 before—a reduction of about 25%, not half.

Those figures cover image output, not the full cost of every API request. Google separately charges for input tokens and text or “thinking” output, and the pricing page lists no free API tier for Nano Banana 2.1. For developers processing large batches, Google also offers a lower-priced Batch option; the image-output component runs $0.0168 at 1K, $0.0252 at 2K and $0.0567 at 4K.

## Image improvements target wide-format artifacts

The new model, identified in the API as `gemini-nano-banana-2.1`, is a stable release and an update to Nano Banana 2, also known as Gemini 3.1 Flash Image. Google says it improves visual quality, prompt adherence, multi-turn character consistency and text rendering, while keeping Flash-level speed and cost efficiency. Its output options include 1K, 2K and 4K resolutions, with 1K as the default.

Google describes the wide and panoramic aspect ratios—1:4, 4:1, 1:8 and 8:1—as existing formats where Nano Banana 2.1 fixes tiling artifacts at 2K and 4K, rather than newly adding those formats. The model can combine up to 14 reference images, maintaining consistency for as many as four characters and fidelity for up to 10 objects. Developers can choose among minimal, medium and high thinking levels; medium is the default.

The model accepts text, image, video and PDF inputs, and returns image and text output. Its listed limits are 131,072 input tokens and 32,768 output tokens. The model page identifies the stable model version as `gemini-nano-banana-2.1`, while Google’s release notes call it generally available and direct developers to migrate from `gemini-3.1-flash-image`.

## Search grounding predates the update

Nano Banana 2.1 supports grounding with Google Web and Image Search, but that capability was also listed for Nano Banana 2. The pricing page lists 5,000 free search requests per month shared across Gemini 3.x models, followed by a charge of $14 per 1,000 requests for text and image-based grounding. That is separate from image output pricing: a customer request may trigger one or more search queries, and Google says each individual query is charged. The page also says retrieved text or images supplied through search grounding are not charged as input tokens.

The cost comparison is therefore useful but narrower than a headline “half-price” claim suggests. The standard image-output charge falls by half at 1K and close to half at 2K, but by roughly a quarter at 4K. Additional request costs vary with inputs, reasoning and search use. Google’s documented price advantage is clearest for developers generating standard-resolution images through the Gemini API.

## Sources

1. Original source [Google AI for Developers · Gemini nano banana 2.1](https://ai.google.dev/gemini-api/docs/models/gemini-nano-banana-2.1) ai.google.dev

Other pages read

2. [Google AI for Developers · page 2](https://ai.google.dev/gemini-api/docs/pricing?authuser=2) ai.google.dev
3. [Google AI for Developers · page 3](https://ai.google.dev/gemini-api/docs/changelog?hl=en) ai.google.dev

## Companies in this story

- [Google](https://ai.info/companies/google)

## Related stories

- [Alibaba Releases Qwen-Image-2.1-Turbo With Eight-Step Sampling](https://ai.info/news/alibaba-qwen-image-2-1-turbo-eight-steps)
- [Qwen-Image 2.1 Released on ModelScope With Transparent Image Generation and Editing](https://ai.info/news/qwen-image-2-1-native-transparent-image-creation)
- [White Circle Finds DALL·E 3 Guardrails Fail in 13% of Attacks](https://ai.info/news/white-circle-dalle-3-guardrails-13-percent-attacks)
- [Bria Releases Fibo 2.0, an 8B Model for Image Editing and Generation](https://ai.info/news/bria-fibo-2-open-8b-image-model)
- [Google Opens SynthID Detector to Everyone Worldwide](https://ai.info/news/google-opens-synthid-detector-worldwide)

## Explore

[More articles](https://ai.info/news)

---

Canonical: https://ai.info/news/google-nano-banana-2-1-lower-image-prices
