Nano Banana 2.1 API Pricing: Token Rates and Migration Costs

Standard Nano Banana 2.1 API pricing is $30.00 per million image-output tokens, which translates to $0.0336 per 1K image, $0.0504 per 2K image, and $0.113 per 4K image. Input tokens cost $1.50 per million across text, image, and video modalities, while text and thinking output tokens cost $7.50 per million.

Google released the model generally available (GA) under the identifier gemini-nano-banana-2.1 on October 6, 2026, marking it as an update to Nano Banana 2 (gemini-3.1-flash-image). The release maintains Flash-level speed while improving visual quality, prompt adherence, multi-turn character consistency, text rendering, and panoramic generation.

Moving workloads to the new model lowers standard image generation costs by roughly half at 1K and 2K resolutions, but it triples prompt input token fees to $1.50 per million and increases text and thinking output fees by 2.5 times to $7.50 per million. The saving on image output has to be weighed against the higher input and text rates.

Nano Banana 2.1 API Pricing per Image and per Token

Nano Banana 2.1 API pricing establishes a standard base rate of $1.50 per million input tokens, $7.50 per million text and thinking output tokens, and $30.00 per million image-output tokens. According to Google's official Gemini API pricing documentation as of October 2026, standard per-image billing is set at $0.0336 for 1K resolution, $0.0504 for 2K resolution, and $0.113 for 4K resolution.

These rates diverge sharply from the older Nano Banana 2 (gemini-3.1-flash-image) rate card. Nano Banana 2 billed input tokens at $0.50 per million and text or thinking output tokens at $3.00 per million, while charging $60.00 per million image-output tokens. On a per-image basis, Nano Banana 2 cost $0.045 at 0.5K, $0.067 at 1K, $0.101 at 2K, and $0.151 at 4K.

Google requires a paid billing account to access the Application Programming Interface (API) for both models. Free tier access is not available for input or output processing on either release.

• Standard input token pricing: $1.50 per million tokens (text, image, and video).
• Standard text and thinking output pricing: $7.50 per million tokens.
• Standard image-output pricing: $30.00 per million tokens.
• Per-image generation cost: $0.0336 at 1K, $0.0504 at 2K, and $0.113 at 4K resolution.

Cost Shift Dynamics for Workload Migration

Migrating an image generation pipeline from Nano Banana 2 to Nano Banana 2.1 cuts raw image-output fees by exactly half while tripling input token costs. The image-output rate drops from $60.00 to $30.00 per million tokens, reducing 1K images from $0.067 to $0.0336 (about 50 percent lower) and 2K images from $0.101 to $0.0504 (about 50 percent lower). At 4K resolution, the cost drops from $0.151 to $0.113, representing a smaller reduction of approximately 25 percent.

The financial trade-off centers on prompt weight and reasoning length. Input tokens cost $1.50 per million on Nano Banana 2.1 compared to $0.50 per million on Nano Banana 2, an increase of $1.00 per million tokens. Text generation and thinking reasoning tokens cost $7.50 per million on 2.1 compared to $3.00 per million on the older version, making long thinking traces 2.5 times more expensive.

Consider an enterprise pipeline producing 10,000 standard 1K images in a monthly billing cycle. In raw image-output fees, Nano Banana 2.1 costs $336.00 (calculated as 10,000 multiplied by $0.0336), whereas Nano Banana 2 incurred $670.00 (calculated as 10,000 multiplied by $0.067). This isolated generation step saves $334.00, though prompt tokens, text responses, thinking traces, and grounding requests are billed separately on top of these figures.

The more input and text or thinking tokens a request carries, the more of the image-output saving the higher rates take back; the lists give only the per-million rates, so test with your own prompts. See the site's separate Nano Banana 2 Lite for business guide for the Lite model.

• 1K resolution output: $0.0336 on 2.1 versus $0.067 on Nano Banana 2 (about 50 percent lower).
• 2K resolution output: $0.0504 on 2.1 versus $0.101 on Nano Banana 2 (about 50 percent lower).
• 4K resolution output: $0.113 on 2.1 versus $0.151 on Nano Banana 2 (about 25 percent lower).
• Input prompt token expense: $1.50 on 2.1 versus $0.50 on Nano Banana 2 (three times).
• Text and thinking output expense: $7.50 on 2.1 versus $3.00 on Nano Banana 2 (2.5 times).

Batch Processing Rates and High-Volume Expenses

Google provides Batch API processing for Nano Banana 2.1 at a 50 percent discount compared to standard rates across all token categories. On the batch tier, input tokens cost $0.75 per million, text and thinking output tokens cost $3.75 per million, and image-output tokens cost $15.00 per million.

Per-image generation fees on the Batch API are priced at $0.0168 for 1K resolution, $0.0252 for 2K resolution, and $0.0567 for 4K resolution. By comparison, Nano Banana 2 batch processing cost $0.022 for 0.5K, $0.034 for 1K, $0.050 for 2K, and $0.076 for 4K.

For high-volume asynchronous jobs, Nano Banana 2.1 batch image output is roughly half the cost of Nano Banana 2 batch generation at 1K ($0.0168 versus $0.034) and 2K ($0.0252 versus $0.050). At 4K resolution, batch generation provides an approximate 25 percent saving ($0.0567 versus $0.076).

• Batch input token rate: $0.75 per million tokens.
• Batch text and thinking output rate: $3.75 per million tokens.
• Batch image-output rate: $15.00 per million tokens.
• Batch per-image generation: $0.0168 at 1K, $0.0252 at 2K, and $0.0567 at 4K.

Model Specifications and Functional Capabilities

Google documents Nano Banana 2.1 under the formal model identifier gemini-nano-banana-2.1 with Stable release status. The architecture accepts text, image, video, and Portable Document Format (PDF) files as input, and it produces image and text outputs.

The model page lists limits of 131,072 input tokens and 32,768 output tokens. Resolution options include 1K, 2K, and 4K, with 1K configured as the default output size. The changelog lists panoramic aspect ratios of 1:4, 4:1, 1:8 and 8:1.

Supported capabilities include image generation, search grounding, Batch API execution, multi-image fusion with up to 14 reference images, and thinking reasoning. Callers can configure thinking levels across three distinct settings: minimal, medium (the default configuration), or high.

Google explicitly lists unsupported features on its official model documentation. Nano Banana 2.1 does not support audio generation, prompt caching, code execution, file search, function calling, Google Maps grounding, the Live API, structured outputs, URL context, Flex inference, or Priority inference.

Migration Protocol and Operational Deprecation Status

According to the Gemini API changelog dated October 6, 2026, gemini-3.1-flash-image is officially deprecated, and Google instructs developers to migrate to gemini-nano-banana-2.1. The changelog announced no retirement or shutdown date for Nano Banana 2. For historical reference, Google's documentation lists October 2, 2026 as the shutdown date for the older Gemini 2.5 Flash Image (gemini-2.5-flash-image), but leaves the Nano Banana 2 shutdown timeline unstated.

Migrating means pointing requests at the gemini-nano-banana-2.1 model code. Because Google updated visual quality, prompt adherence, multi-turn character consistency, text rendering, and panoramic outputs, engineering teams must re-test prompts to verify that generated imagery satisfies production requirements.

Grounding costs remain separate from token usage. Grounding with Google Web and Image Search includes 5,000 free requests per month shared across Gemini 3 and newer foundation models. Beyond that monthly allowance, Google bills search grounding at $14.00 per 1,000 requests.

Documentation Gaps and Evaluation Conditions

Several operational parameters remain unpublished in Google's official documentation. Google does not publish a knowledge cutoff date for Nano Banana 2.1, nor does it list an exact release timestamp beyond the October 6, 2026 changelog publication date.

Google also publishes no per-image rate for 0.5K resolution on Nano Banana 2.1, providing rate rows only for 1K, 2K, and 4K dimensions. Google provides no numerical performance benchmarks, latency metrics, or hardware speed scores for the new model, describing its performance qualitatively as high-efficiency with Flash-level speed.

Google's documentation does not publish whether Nano Banana 2 historically supported the specific tools now marked unsupported on 2.1, such as structured outputs or function calling. Developers must inspect their existing codebases against the unsupported capabilities list, calculate prompt token weight, and check current figures on Google's pricing documentation before executing full production cutovers.