Google has released Nano Banana 2.1, a new image-generation and editing model that is being added to the Gemini app, Google Search’s AI Mode, Google Ads and a range of developer products. The company says the model improves on earlier versions while reducing the cost of generating images through its API.
What Google says has improved
Nano Banana 2.1 converts written prompts into images and can also modify existing photographs. Google highlights three main upgrades: stronger visual design, more precise mask-based editing and improved subject consistency. Mask-based editing allows users to identify a specific part of an image and change that area without altering the rest. Subject consistency is intended to help people and objects retain their appearance through multiple edits.
Google’s internal testing gave Nano Banana 2.1 an overall text-to-image preference score of 1,050 ELO points. Nano Banana 2 scored 990, while Nano Banana Pro recorded 935. ELO is a comparative scoring system in which higher results indicate stronger performance; the score does not have a fixed maximum.
The company also reported an infographic factuality score of 0.521 for the new model, compared with 0.179 for Nano Banana 2. However, independent testing was not available when the model launched, so Google’s results have not yet been externally verified.
More references and cheaper API access
According to Decrypt, Nano Banana 2.1 can handle as many as 14 reference images in a single task, while tracking up to four characters and 10 objects. It can produce images at resolutions of up to 4K and in formats with aspect ratios as wide as 8:1.
Developers can choose how much processing the model performs before producing an image, from minimal to high. They can also allow it to use Google Search and Google Image Search as sources when generating an image, enabling the model to look up information about real-world subjects first.
The developer API price for a standard 1K image starts at $0.0336, compared with $0.067 for Nano Banana 2. A 4K image costs $0.0756, down from $0.151. Batch processing, which handles groups of images rather than producing them immediately, receives an additional 50% discount.
At the standard 1K rate, 1,000 images would cost about $33.60 through the API, versus $67 using Nano Banana 2. The lower price could make the model less expensive for applications that generate images at scale, although the launch material does not provide independent performance comparisons.
Why the model matters for Google
Nano Banana 2.1 follows the release of version 2 in February, which was built on Gemini 3.1 Flash Image and introduced the ability to check Google Search before creating images. The original Nano Banana gained broad attention in September 2025 after its selfie-to-collectible-figurine feature helped Gemini reach No. 1 on both major app stores, ahead of ChatGPT. Alphabet’s market value also exceeded $3 trillion during that period.
Moneyiar market picture
Moneyiar data at 10:24 a.m. on October 7, 2026, showed FORM at $0.2703153976, down 3.39% over 24 hours and 6.79% over seven days. DEL was $0.3867052073, up 0.70% in 24 hours but down 7.98% over seven days. BAN stood at $0.0638205769, falling 5.64% daily and 0.39% weekly, while XNO was $0.3502553211, down 3.58% daily and 4.15% over seven days.
