Google Launches Nano Banana 2.1 AI Model
Google has launched Nano Banana 2.1, its latest image generation and editing model. The model became generally available on October 6, 2026.
The update focuses on visual quality, editing precision and subject consistency. In addition, Google has improved prompt adherence and text rendering. The model is based on Gemini 3.6 Flash and is optimised for high-efficiency image generation.
Moreover, Nano Banana 2.1 supports image generation at 1K, 2K and 4K resolutions. It also introduces configurable Thinking levels for more complex image tasks.
Stronger Subject Consistency and Editing
One of the main upgrades involves maintaining subjects across multiple images. Nano Banana 2.1 can preserve the appearance of up to five characters in a workflow. It can also maintain fidelity for up to 14 objects.
Consequently, users can create visual sequences with more consistent characters and objects. This could benefit storytelling, advertising, product visualisation, and other creative workflows.
The model also improves mask-based and local editing. Therefore, users can make targeted changes while retaining more of the surrounding image. Google’s evaluations show gains in general editing, character consistency and mask-based editing.
Furthermore, Nano Banana 2.1 supports multi-image fusion. Developers can combine multiple reference images in a single workflow to guide the final result.
Google also says the model improves visual realism and text rendering. In addition, it addresses tiling artefacts in wide and panoramic images at higher resolutions.
More References and Better Visual Control
Nano Banana 2.1 can process up to 14 reference images. However, Google specifies that object and character fidelity limits vary within that broader reference-image capability.
The model also integrates grounding from Google Web and Image Search. As a result, image creation can draw on current information and visual references when those capabilities are enabled.
Meanwhile, developers can choose between minimal, medium and high Thinking levels. The medium setting serves as the default.
This gives users more control over the balance between reasoning depth and generation speed. Therefore, the model can support both quick creative tasks and more demanding visual workflows.
Google’s own evaluations show improvements over its earlier image models. Nano Banana 2.1 scored higher in general editing, multi-character consistency and mask-based editing tests than Nano Banana 2 and Nano Banana Pro in the company’s reported comparisons.
Broad Rollout Across Google Platforms
Nano Banana 2.1 is rolling out across several Google products. These include the Gemini app, Google AI Studio, Gemini API, Google Search AI Mode, Google Ads, Flow and Stitch.
For developers, Google identifies gemini-nano-banana-2.1 as the stable model. The company positions it as a high-efficiency option alongside its more capable Nano Banana Pro model.
The model also supports a context window of up to 1 million tokens. Additionally, it can accept text and image inputs while producing image and text outputs.
The rollout comes as image generation becomes increasingly important across creative and commercial applications. Therefore, stronger consistency and precise editing could make AI-generated imagery more practical for repeated production workflows.
However, Google notes that Nano Banana 2.1 still has limitations. These include occasional spatial confusion and constraints in advanced world knowledge, 3D reasoning and factuality.
Even so, the release strengthens Google’s position in AI image generation. With improved editing, reference handling, and subject consistency, Nano Banana 2.1 provides developers and creators with a more capable Flash-based image model.
