Google Unveils Nano Banana 2.1 with Advanced Editing Tools and Lower API Pricing

Google has officially released Nano Banana 2.1, the latest iteration of its flagship artificial intelligence model designed to convert typed text prompts into vivid imagery or perform intricate edits on pre-existing photographs. Building upon the widespread success of its predecessors, the new model aims to push the boundaries of AI-driven visual design while simultaneously introducing significant performance upgrades and cost reductions for both everyday users and enterprise developers alike.

The rollout of Nano Banana 2.1 is already underway across a robust suite of Google ecosystems. Users can access the new model directly within the Gemini application, Google Search’s dedicated AI Mode, and Google Ads. Additionally, the technology is being integrated into key developer platforms and creative environments, including Google AI Studio, Flow, and Stitch, ensuring that creators and engineers can leverage the upgraded capabilities across multiple workflows.

The original Nano Banana model played a pivotal role in catapulting Gemini to mainstream prominence. In September 2025, a viral feature that allowed users to transform casual selfies into collectible figurines helped drive Gemini to the number one spot on major app stores, effectively dethroning ChatGPT after a nearly three-year reign at the top. The immense popularity of the tool also coincided with a period of explosive financial growth for parent company Alphabet, which saw its market capitalization surpass the $3 trillion threshold around the same timeframe.

Following that breakthrough, Google introduced Nano Banana 2 in February, powered by the Gemini 3.1 Flash Image architecture. That version introduced a vital capability: the model could cross-reference Google Search before rendering real-world objects, significantly improving the accuracy of images depicting actual historical events, public figures, or specific locations.

With the arrival of Nano Banana 2.1, Google is focusing heavily on refining visual fidelity, precision, and efficiency. According to official announcements from the company, the upgraded model outperforms all previous iterations across the board. Google specifically highlighted notable leaps in overall visual design quality, sharper mask-based editing capabilities, and enhanced subject consistency. The latter improvement ensures that specific characters or objects retain their distinct identity and features even after undergoing multiple rounds of successive edits.

To quantify these advancements, Google pointed to benchmark testing using the ELO rating system, where Nano Banana 2.1 achieved a score of 1,050 points. For comparison, the preceding Nano Banana 2 scored 990 points, while the specialized Nano Banana Pro recorded 935 points. In the ELO qualitative framework, higher scores reflect a superior preference among users during blind tests, indicating a clear trajectory of continuous improvement in the model’s output quality.

Under the hood, Nano Banana 2.1 boasts impressive technical specifications. On paper, the architecture is capable of handling up to 14 reference images simultaneously. Within that data stream, it can effectively track up to four distinct characters and ten individual objects, maintaining coherence across complex scenes. The model is capable of outputting visuals in resolutions up to 4K—approaching standard television sharpness—and can generate content in ultra-wide aspect ratios reaching up to 8:1.

Developers working with the model are also given granular control over the generation process. Through the system settings, they can dictate how long the AI "thinks" before rendering an image, scaling the processing time from minimal to high depending on the complexity of the task. Furthermore, developers can enable grounding features that allow the model to consult Google Search and Google Image Search dynamically, ensuring that generated graphics remain anchored in accurate, real-world data.

Alongside these creative and technical enhancements, Google has restructured its pricing model through the official developer API, making high-end AI image generation substantially more accessible. For a standard 1K resolution image, the API cost has been slashed to $0.0336, representing a fifty percent reduction compared to Nano Banana 2, which previously cost $0.067 per image.

Scaling up to higher resolutions yields similar savings. A 4K image now runs $0.0756 through the developer pipeline, down significantly from the previous cost of $0.151. Furthermore, enterprise users and developers executing batch jobs—processing large volumes of images in bulk rather than requesting instantaneous single generations—qualify for an additional fifty percent discount.

At the standard 1K pricing tier, generating a batch of one thousand images now amounts to roughly $33.60 via the API, a steep drop from the $67 price tag associated with Nano Banana 2. By combining these dramatic cost efficiencies with sharper editing controls, higher resolution outputs, and deeper integration across Google’s suite of consumer and developer products, Nano Banana 2.1 establishes a new benchmark for generative imaging technology.