Google is making its latest image model cheaper to run while giving developers more control over how subjects and objects survive repeated edits.
The company released Nano Banana 2.1 on Tuesday, an image generation and editing model built on Gemini 3.6 Flash. It is rolling out across the Gemini app, Google AI Studio, Gemini API, Google Search’s AI Mode, Google Ads, Flow and Stitch.
The model accepts text and images, supports up to 1 million tokens of context, and can generate images at 1K, 2K and 4K resolutions. Google says it improves visual quality, text rendering, mask-based editing and consistency when the same subjects are edited repeatedly.
One of the bigger upgrades is multi-image editing. Nano Banana 2.1 can process up to 14 reference images, while maintaining consistency for up to four characters and fidelity for up to 10 objects, according to Google.
The price cut could matter more than the pixels
For developers, the launch is notable because Google has cut image-output pricing substantially. Nano Banana 2.1 costs $30 per million image tokens on the standard paid Gemini API tier, compared with $60 for Nano Banana 2.
That works out to about $0.0336 for a 1K image, $0.0504 for 2K and $0.113 for 4K output. Batch processing cuts those image costs in half again, according to Google’s pricing information. The savings could make repeated image generation more practical for applications such as advertising, product visualization and content production, where thousands of generations can quickly become expensive.
The most important part of Nano Banana 2.1 may not be its 4K output. It’s Google’s attempt to make high-quality image generation cheap enough to become a routine part of software products.
Google’s benchmarks show a sizable jump
Google’s own testing gives Nano Banana 2.1 a 1,050 Elo score for overall text-to-image preference in its Thinking configuration, compared with 990 for Nano Banana 2 and 935 for Nano Banana Pro. In Google’s evaluation system, a higher Elo score indicates that testers preferred the model more often in head-to-head comparisons.
The model also scored higher across Google’s listed editing evaluations, including multi-character consistency, stylization, product consistency and multi-reference editing. Those numbers should still be viewed in context. They are Google’s evaluations, rather than independent testing, so they show where Google believes the model has improved rather than proving it will win every real-world comparison.
More Google coverage
- New Google Search AI Mode is ‘Total Reimagining,’ Says CEO Sundar Pichai
- In Major Ruling, Judge Finds Google ‘Willfully Acquired and Maintained Monopoly Power’ Over Digital Ad Market
- Google’s Big Bet on Nuclear Energy: ‘The Race to Power AI-Driven Data Centers is Accelerating’
- Computer History Museum Releases Original AlexNet Code: Why It Matters
What the update means for creators and developers
Nano Banana 2.1 is particularly interesting for workflows that involve several rounds of editing. Better subject consistency means creators can make changes without having a character, product or other visual element drift dramatically between generations.
Search grounding could also help with images that depend on real-world information. But it does not remove the need to check the result. Google’s model card warns of hallucinations, occasional timeouts, weak small-text rendering, imperfect spatial localization and limitations in world knowledge, 3D reasoning and factuality.
Google also says some domains may have knowledge limited to January 2025 despite a March 2026 cutoff for Gemini 3.6 Flash. For enterprise teams and API builders, however, the calendar demands attention.
With Nano Banana 2 slated for deprecation by late October, developers have roughly three weeks to migrate endpoints and benchmark output quality against existing production workloads.
The upgrade therefore comes with both an incentive and a deadline. Lower image-generation costs make Nano Banana 2.1 easier to justify at scale, but teams already running Nano Banana 2 should use the remaining migration window to test output quality, consistency and cost before switching production traffic.
Want to learn more AI tips, tricks, and prompting techniques? TechRepublic readers get free 7-day access to The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work. Browse all lessons →
Let us teach you How to Talk to AI for free! Try our six-minute course at The Neuron Academy and learn a few simple ways to write better prompts and get more useful results from AI, or browse our other AI course for free for seven days. Check out all the lessons here →