Google's Nano Banana 2.1 tops its own Pro model at half the image cost

Nano Banana 2.1 arrives as the successor to Nano Banana 2, which Google released in February and which is also known as Gemini 3.1 Flash Image, so this is a refresh roughly eight months after that launch. The model card says it is built on Gemini 3.6 Flash, supports text and image inputs and offers a context window of up to 1 million tokens. Google's own comparisons put it above both the model it replaces and its higher-tier sibling, Nano Banana Pro, which is also called Gemini 3 Pro Image — on human-scored Elo for text-to-image generation, on automatic scoring of infographic factual accuracy, and on all seven evaluated image-editing items, including consistency across multiple characters and edits guided by masks or hand-drawn instructions.

Google also set out the limits it has not solved. Small text and long paragraphs can come out blurry, character consistency between an input and a generated image is not always perfect, and the model can mix up spatial relations such as left and right. Those caveats sit alongside an unusual pricing move: image output was cut to roughly half of the previous model's per-image cost, but input pricing rose from $0.50 to $1.50 per 1 million tokens, and the API has no free tier. Developers who generate many images with short prompts are likely to benefit most, while those who send large inputs may find the balance less favorable, though the actual cost depends on each workload's mix of input, text output and images.