Google releases Nano Banana 2.1 for image creation and editing
The new Gemini image model is documented as stable, with updated editing controls and a lower listed API cost for a 1K image.
A stable image model with new controls
Google has released Nano Banana 2.1, a Gemini image generation and conversational editing model built on Gemini 3.6 Flash. Its API documentation lists the stable model code gemini-nano-banana-2.1 and support for 1K, 2K and 4K image output. Google highlights improved text rendering, visual quality, multi-turn character consistency and editing across multiple reference images. The documentation also lists adjustable thinking levels and grounding with Google Web and Image Search. The Decoder reported that the model is rolling out across Google products, including the Gemini app, AI Mode and developer tools; that rollout may vary by product and user.
Price and performance need context
Google’s current API pricing table lists a standard 1K output image at $0.0336, compared with $0.067 for the preceding Nano Banana 2 model. That is a comparison of listed per-image output prices, not a promise that every request or workflow costs half as much. Google reports stronger results in its model-card evaluations, but those figures come from its own tests. The model card also notes persistent limitations: small text can be blurry, character consistency is imperfect, and masked edits can partly miss instructions. The Decoder’s hands-on comparison found that Nano Banana Pro still looked more natural in one image example, illustrating why vendor benchmarks should not be treated as universal rankings.