Nano Banana 2.1 debuts at half the image cost of its predecessor: Unite.AI

Google released Nano Banana 2.1, an image generation and editing template based on Gemini 3.6 Flash, on October 6, 2026, with availability on the Gemini app, Google AI Studio, Gemini API, and Gemini Enterprise Agent Platform, and paid tier image output API priced at $30 per million tokens.

The Nano Banana 2.1 model sheet, released on the same day, describes the model as a member of the Gemini 3 series of natively multimodal reasoning models. It accepts text and image strings as input with a context window of up to 1 million tokens and generates output of images and text, listed with 4K tokens and 64K tokens respectively. The tab lists distribution via the Gemini app, Google AI Studio, Gemini API, AI mode in Google Search, Google Ads, Google Flow, and Google Stitch.

Enterprise platform specifications

On the Gemini Enterprise Agent Platform, the model carries the identifier gemini-nano-banana-2.1 in the generally available launch phase, with a release date listed as October 6, 2026. Google’s platform documentation describes Nano Banana 2.1 as optimized for multimodal image generation and editing, offering a balance between price and performance.

The documentation lists text and images as supported for input and output, video as input-only, and audio as unavailable, with a 131,072-token context window and a maximum of 32,768 output tokens. The seed, topK, logprobs, temperature, and topP parameters are not supported, and setting any of them returns an API error.

Supported features include reasoning, system instructions, implicit context caching, token counting, rooting with Google Search and image search, provisioned throughput, batch inference, and standard pay-as-you-go, with global availability. Image generation, including from video input, image editing and multi-round editing, interlaced images and text, content credentials (C2PA) and virtual proof are supported; generation of people is not available.

Limits include 14 images per prompt, 7 MB per file for inline data or console uploads, 30 MB per file from Cloud Storage, and a 500 MB input size limit. Supported aspect ratios include 1:1, 3:2, 4:3, 16:9, 9:16, and 21:9, with 1K, 2K, and 4K resolutions, and support for png, jpeg, webp, heic, and heif files. The documentation states that the model consumes 1,120 tokens per input image, while the output images consume 1,120 tokens at 1K resolution and 1,680 tokens at 2K resolution.

Gemini API pricing

Google’s Gemini API pricing page, updated on October 6, 2026, describes Nano Banana 2.1 as “an update to Nano Banana 2 (Flash Gemini 3.1 image),” built for high-efficiency image generation and conversational editing with improved visual quality, multi-turn font consistency, accurate text rendering, and search-based generation at 1K, 2K, and 4K resolutions. The standard paid price is listed at $1.50 per million input tokens for text, image, and video, $7.50 per million output tokens for text and thought, and $30.00 per million tokens for image output, which the page equates to $0.0336 per 1K image, $0.0504 per 2K image, and $0.0756 per 4K image.

Batch pricing is listed at $0.75 per million input tokens, $3.75 per million tokens for text and thought output, and $15.00 per million image tokens, with stated equivalencies of $0.0168, $0.0252, and $0.0378 per 1K, 2K, and 4K image respectively. Grounding with Google web and image search results in 5,000 free requests per month shared across Gemini 3.x models, so $14 per 1,000 requests. The page does not list free tier access for Nano Banana 2.1.

The same page lists its predecessor, Gemini 3.1 Flash Image (Nano Banana 2), at $0.50 per million input tokens, $3 per million tokens for text and thought output, and $60.00 per million image tokens, with per-image equivalencies of $0.067 at 1K, $0.101 at 2K, and $0.151 at 4K.

Ratings reported

The model sheet reports a rating approach that combines side-by-side human ratings that produce Elo scores through text-to-image conversion and editing tasks, a one-sided AutoRater for factuality, and regression sets from the Gemini 2.5 Flash Image and Gemini 3 Pro Image use cases.

As for text-to-image, the card reports a general preference Elo of 1050 ±14 for Nano Banana 2.1 in its Thinking configuration and 1015 ±13 without thinking, against 990 ±7 for Nano Banana 2 (Thinking) and 935 ±8 for Gemini 3 Pro Image (Nano Banana Pro). Infographic Design scores reported are 1048 ±17 for the Thinking configuration, versus 961 ±12 for Nano Banana 2 and 912 ±12 for Nano Banana Pro, while Infographic Factuality is listed at 0.521, versus 0.179 and 0.265.

For editing in the Thinking configuration, the tab lists Nano Banana 2.1 at 1026 for overall editing, 1028 for single character consistency, 1106 for multiple character consistency, 1049 for mask/ink editing, 1024 for product consistency, 1062 for stylization, and 1066 for multiple reference editing, each listed above the corresponding scores of Nano Banana 2 and Nano Banana Pro in the same table.

Limitations and safety assessments

The fact sheet lists known limitations including hallucinations, occasional slowness or timeouts, poor rendering of small text and long paragraphs, imperfect character consistency between input and generated images, partial following of instructions and ink persistence in masked editing, rare cases of persistent subject pose during editing, occasional confusion over spatial location, and limited knowledge of the world, 3D reasoning, and factuality. Gemini 3.6 Flash has an awareness limit of March 2026, although the board notes that in some domains awareness may be limited to January 2025.

The sheet reports manual red teaming by teams of specialists external to the model development team, states that Nano Banana 2.1 met the required launch thresholds for child safety, and describes its safety performance as similar or improved compared to Gemini 3 Flash, without finding serious concerns compared to Gemini 3.1 Pro. For border security, the sheet states that Gemini 3.1 Pro and Gemini 3.7 Flash have not reached any charted or critical capability level and that Nano Banana 2.1 shows no significant new capabilities or material performance increases over those models, leaving it deemed unlikely to reach those levels.