Google announced Nano Banana 2.1, an AI model for generating and editing images, on October 6 local time. The update to February’s Nano Banana 2, also called Gemini 3.1 Flash Image, began rolling out that day across the Gemini app, AI Mode in Google Search, Google AI Studio, Flow, Stitch, Google Ads, and Gemini Enterprise Platform.

Google said Nano Banana 2.1 outperforms its predecessor across all areas, with particular gains in visual design, mask-based editing, and subject consistency. A model card released the same day says the model is based on Gemini 3.6 Flash, accepts text and image inputs, and has a context window of up to 1 million tokens.

In Google’s Elo scores, calculated from human comparisons, Nano Banana 2.1 scored 1,050 overall for text-to-image generation with Thinking enabled. Nano Banana 2 scored 990, while the higher-tier Nano Banana Pro (Gemini 3 Pro Image) scored 935. In an automated test of factual accuracy in infographics, Nano Banana 2.1 scored 0.521, compared with 0.179 for Nano Banana 2 and 0.265 for Nano Banana Pro.

For image editing, Nano Banana 2.1 scored 1,106 for consistency across multiple characters, ahead of Nano Banana 2 at 978 and Nano Banana Pro at 1,011. It scored 1,049 for edits based on masks or handwritten instructions, compared with 965 and 927 for the two other models. Although it beat both models in all seven evaluated tasks, Google listed several areas for improvement: small text and long paragraphs can appear blurry; characters may not remain consistent between input and generated images; and the model may confuse spatial relationships such as left and right.

In Google apps, users can access the model in Search AI Mode through the banana icon below the search box. Developers can use it through the Gemini API under model ID “gemini-nano-banana-2.1”; it is available only on paid tiers. Pricing per 1 million tokens is $1.50 for input, $7.50 for text and thinking output, and $30 for image output.

Per-image output costs $0.0336 at 1K (1024 × 1024 pixels), $0.0504 at 2K (2048 × 2048 pixels), and $0.0756 at 4K (4096 × 4096 pixels). Nano Banana 2 costs $0.067, $0.101, and $0.151 at the same resolutions, making Nano Banana 2.1’s image output nearly half as expensive. Input pricing rose from $0.50 for Nano Banana 2 to $1.50. All prices are halved when using the Batch API.

Google announced Nano Banana 2.1, an AI model for generating and editing images, on October 6 local time. The update to February’s Nano Banana 2, also called Gemini 3.1 Flash Image, began rolling out that day across the Gemini app, AI Mode in Google Search, Google AI Studio, Flow, Stitch, Google Ads, and Gemini Enterprise Platform.

Google said Nano Banana 2.1 outperforms its predecessor across all areas, with particular gains in visual design, mask-based editing, and subject consistency. A model card released the same day says the model is based on Gemini 3.6 Flash, accepts text and image inputs, and has a context window of up to 1 million tokens.

In Google’s Elo scores, calculated from human comparisons, Nano Banana 2.1 scored 1,050 overall for text-to-image generation with Thinking enabled. Nano Banana 2 scored 990, while the higher-tier Nano Banana Pro (Gemini 3 Pro Image) scored 935. In an automated test of factual accuracy in infographics, Nano Banana 2.1 scored 0.521, compared with 0.179 for Nano Banana 2 and 0.265 for Nano Banana Pro.

For image editing, Nano Banana 2.1 scored 1,106 for consistency across multiple characters, ahead of Nano Banana 2 at 978 and Nano Banana Pro at 1,011. It scored 1,049 for edits based on masks or handwritten instructions, compared with 965 and 927 for the two other models. Although it beat both models in all seven evaluated tasks, Google listed several areas for improvement: small text and long paragraphs can appear blurry; characters may not remain consistent between input and generated images; and the model may confuse spatial relationships such as left and right.

In Google apps, users can access the model in Search AI Mode through the banana icon below the search box. Developers can use it through the Gemini API under model ID “gemini-nano-banana-2.1”; it is available only on paid tiers. Pricing per 1 million tokens is $1.50 for input, $7.50 for text and thinking output, and $30 for image output.

Per-image output costs $0.0336 at 1K (1024 × 1024 pixels), $0.0504 at 2K (2048 × 2048 pixels), and $0.0756 at 4K (4096 × 4096 pixels). Nano Banana 2 costs $0.067, $0.101, and $0.151 at the same resolutions, making Nano Banana 2.1’s image output nearly half as expensive. Input pricing rose from $0.50 for Nano Banana 2 to $1.50. All prices are halved when using the Batch API.