Nano Banana 2 is Google's stable image generation/editing model with 0.5K–4K output, Search grounding and pricing from about $0.045 per image.
Nano Banana 2 is Google's high-efficiency production image model for fast generation and conversational editing up to 4K, with Search grounding and multimodal reference inputs.
Nano Banana 2 is Google’s production-scale Gemini image model, officially exposed as gemini-3.1-flash-image and positioned as the high-efficiency counterpart to the more expensive Nano Banana Pro.
The model’s main value is not simply that it generates images quickly. Google designed it for iterative production workflows: developers can provide text, existing images, video or PDF context, generate a visual, then continue editing conversationally. The stable version adds multiple output resolutions, improved aspect-ratio adherence, international text rendering and Search/Image Search grounding so generation can incorporate current web information.
Nano Banana 2 accepts up to 131,072 input tokens and has a 32,768-token output limit for its multimodal interaction. It produces both images and text. Supported image resolutions include 0.5K, 1K, 2K and 4K, and Google added extreme aspect ratios such as 1:8 and 8:1 alongside conventional formats. Search grounding can use both text and image results, which can be helpful for current products, places or visual references.
A marketing system can ingest a product brief, brand assets and reference images, then create 1K social variants and 4K campaign assets. A publisher can supply video as context and ask the model to create a representative thumbnail or infographic. Localization teams can iterate on visual content with non-English text, while e-commerce platforms can edit product settings without rebuilding a scene manually.
Nano Banana 2 became generally available on May 28, 2026 after its February preview. Google lists it as Stable and recommends it as the go-to all-around image model for balancing intelligence, cost and latency. The older preview endpoint was deprecated after GA, so new integrations should use gemini-3.1-flash-image rather than the preview ID.
Pricing is unusually transparent. Standard output costs are approximately $0.045 for a 0.5K image, $0.067 for 1K, $0.101 for 2K and $0.151 for 4K. Text/image input is billed at $0.50/M tokens, while text/thinking output has separate token pricing. Batch output images cost about half of standard. Search grounding may add query charges after Google’s monthly free allowance.
Choose Nano Banana 2 for production applications that need high throughput, editing and flexible resolution. Choose Nano Banana Pro when professional layouts or maximum visual quality outweigh cost, and Nano Banana 2 Lite when ultra-low latency and the lowest image price are the primary constraints. For teams already building on Gemini, the shared API ecosystem is an additional advantage.
Even strong image models can miss exact text, alter identity, invent product details or produce inconsistent geometry. Search grounding improves access to current visual information but does not guarantee factual reproduction. Teams should compare generated assets against source material, especially for brand, product, medical, legal or factual imagery, and account for regeneration rates when estimating real cost.
In addition, its main capabilities include Text-to-image, Image editing, 0.5K/1K/2K/4K output, Conversational editing, Video-to-image context and Search grounding. For example, common use cases include Marketing, E-commerce, Design automation, Thumbnails, Localization and High-volume creative production.
In practice, this model may suit High-volume image generation, Image editing, Marketing assets, Product visuals, Search-grounded graphics and Multilingual typography. Also, notable strengths include Stable GA model, Transparent per-resolution pricing, 4K support and Search grounding. However, review trade-offs such as Generated typography and fine details can still fail, Grounding queries can add cost and Not every creative style will preserve source identity perfectly before adopting it.
Meanwhile, Standard paid pricing: $0.50/M text-image input tokens; image output is $0.045 for 0.5K, $0.067 for 1K, $0.101 for 2K and $0.151 for 4K. Batch image prices are approximately half. A standard 1K output costs about $0.067, while 0.5K–4K images range from roughly $0.045 to $0.151; Batch jobs approximately halve image-output cost.
Use the official model website, official documentation, pricing or release source and additional primary source to confirm current availability, limits and pricing. Product details can change after publication, so rely on primary documentation for final decisions.
Next, continue your research in the AI models directory, Google models and AI Image Generation Model models. Compare providers, pricing, modalities and practical limitations side by side to choose the right model for your workflow.
Create a 4K product campaign image from these reference photos while preserving brand colors.Turn this video frame into a polished thumbnail with accurate multilingual text.Edit this image conversationally to change the environment without changing the main subject.Standard paid pricing: $0.50/M text-image input tokens; image output is $0.045 for 0.5K, $0.067 for 1K, $0.101 for 2K and $0.151 for 4K. Batch image prices are approximately half.
One of the most practical current image APIs for production scale because it combines transparent pricing, fast generation, editing and 4K output.
