Google’s Nano Banana 2.1: image controls and documented limits

Google describes Nano Banana 2.1 as an image-generation and conversational-editing model, and says it keeps Flash-level speed and cost efficiency while improving visual quality, prompt adherence, multi-turn character consistency and text rendering.[2] Its API model identifier is `gemini-nano-banana-2.1`.[2] The documentation says visual quality and realism improvements apply across 1K, 2K and 4K output, with 1K selected by default.[2] For wide and panoramic aspect ratios, Google says the model fixes tiling artifacts at 2K and 4K.[2]

For workflows that depend on specific assets, Google says Nano Banana 2.1 can fuse up to 14 reference images, maintain consistency for up to four characters and preserve fidelity for up to 10 objects.[2] The model can also ground its work using Google Web and Image Search.[2] Users can configure thinking as minimal, medium or high, with medium as the default.[2] These are stated capabilities and limits; teams evaluating a workflow can use them to identify whether their reference-image, character and object requirements fit within the documented bounds.

The reported price for a 1K image is 3.36 cents, down from 6.70 cents; a 4K image is 7.56 cents, down from 15.10 cents.[1] The Decoder reported that a 1K Nano Banana Pro image costs 13.40 cents.[1] Google describes Nano Banana 2.1 as the more efficient counterpart to Nano Banana Pro.[2] The Decoder reported that 2.1 sometimes outperformed Pro by a wide margin in benchmarks, but also said Nano Banana Pro often still produces noticeably better images in practice.[1] Google plans to shut down Gemini 3.1 Flash Image on October 29, 2026.[1]