Nano Sub-Second Inference
Optimized for TTFT under 200ms, enabling real-time vision agents and responsive live-camera loops.
Nano Banana 2 (Gemini 3.1 Flash Image)
The Nano banana api provides developers with sub-second visual reasoning and a 1M token context window. Optimized for high-throughput OCR, native video analysis, and spatial detection, this API delivers pro-tier vision at a fraction of the cost.
API Usage Examples
curl --request POST "https://gptproto.com/api/v3/google/gemini-3.1-flash-image/text-to-image" \
--header "Authorization: Bearer $GPTPROTO_API_KEY" \
--header "Content-Type: application/json" \
--data '{
"prompt": "A tiny origami fox sailing a teacup across a moonlit puddle",
"aspect_ratio": "16:9",
"size": "1k",
"output_formart": "png",
"enable_base64_output": false,
"enable_sync_mode": false
}'Nano Banana 2 (Gemini 3.1 Flash Image) pricing
Start from the cost of a single sample and pick a testing budget. GPTProto rates are 40% below list price.
Your usage
$0.0402 / image
Related Models
All Models
| Model | Resolution | Input → Output | ||
|---|---|---|---|---|
| $0.04 per time | 1K · 2K · 4K | Input: TextInput: Image Output: Image | |
| $7.60 / $28.50 per 1M | — | Input: TextInput: Image Output: Image | ||
| $7.60 / $28.50 per 1M | — | Input: TextInput: Image Output: Image | ||
| — | $0.04 per time | 1K · 2K · more | Input: TextInput: Image Output: Image | |
| $0.04 per time | 1K · 2K | Input: TextInput: Image Output: Image | ||
| $0.02 per time | 1K | Input: TextInput: Image Output: Image | ||
| $6.40 / $24.00 per 1M | — | Input: TextInput: Image Output: Image | ||
| $0.04 per time | 1K · 2K · 4K | Input: TextInput: Image Output: Image | ||
| $0.03 per time | — | Input: TextInput: Image Output: Image | ||
| $0.03 per time | — | Input: TextInput: Image Output: Image | ||
| $0.01 per time | — | Input: TextInput: Image Output: Image | ||
| $0.02 per time | — | Input: TextInput: Image Output: Image | ||
| — | $0.14 per time | — | Input: Text Output: Image | |
| $0.08 per time | 1K · 2K · 4K | Input: TextInput: Image Output: Image | ||
| — | $0.01 per time | — | Input: Image Output: Image | |
| — | $0.02 per time | — | Input: Image Output: Image | |
| $0.03 per time | — | Input: TextInput: Image Output: Image | ||
| $0.02 per time | — | Input: TextInput: Image Output: Image | ||
| $0.03 per time | — | Input: Text Output: Image | ||
| $0.06 per time | — | Input: TextInput: Image Output: Image | ||
| $0.03 per time | — | Input: TextInput: Image Output: Image | ||
| $0.05 per time | — | Input: Image Output: Image | ||
| $0.05 per time | — | Input: Image Output: Image | ||
| $0.05 per time | — | Input: Text Output: Image | ||
| $0.06 per time | — | Input: TextInput: Image Output: Image |
Core Features of the Nano Banana API
Technical standout features that define the Nano banana api performance.
Nano Video Reasoning
Analyze up to 60 minutes of video natively without the overhead of manual frame extraction.
Nano Context Caching
Store large image sets in KV cache to reduce input costs by up to 90% on frequent requests.
Nano Spatial Detection
Native ability to return normalized bounding box coordinates for objects detected in any image.
Nano Banana API Frequently Asked Questions
Find technical answers regarding the Nano banana api, its vision capabilities, and integration through GPTProto.
How fast is the Nano banana api for vision?
The Nano banana api is engineered for speed, delivering a Time-To-First-Token (TTFT) under 200ms for single-image prompts. This sub-second inference makes it an ideal choice for live-camera agentic loops and real-time visual feedback systems. Unlike heavier models that struggle with latency, this API ensures that visual reasoning happens almost instantaneously, allowing for seamless integration into interactive consumer applications.
Can the Nano banana api analyze long videos?
Yes. The Nano banana api supports native video context, allowing it to process up to 60 minutes of video in a single prompt. It treats video frames as temporal tokens, which eliminates the need for developers to manually extract frames. This makes the Nano banana api highly effective for creating cinematic posters, summary infographics, or performing security feed analysis over long durations without losing context.
What is Nano Context Caching?
Nano Context Caching allows you to store frequently used image sets or massive documentation in the model's KV cache. By using this feature via the Nano banana api, developers can achieve up to a 90% cost reduction on repeat calls. It is particularly useful for Visual RAG applications where the same technical diagrams or blueprints are queried repeatedly, ensuring high performance while maintaining a very low operational budget.
Does this API support structured data extraction?
The Nano banana api excels at high-density OCR and structured data tasks. It offers native support for JSON mode via the response_format parameter. In technical tests, the Nano banana api demonstrates over 98% character accuracy when converting high-resolution PDF scans or handwritten notes into Markdown tables or JSON objects. It is a powerful tool for digitizing millions of invoices or insurance forms at scale.
Is my data private when using the Nano banana api?
Privacy is a core pillar of our platform. Under our enterprise agreement with Google, no data sent through the Nano banana api on GPTProto is used to train foundation models. Your images, videos, and text prompts remain isolated and secure. This ensures the Nano banana api is compliant with strict enterprise requirements for sensitive document processing, medical scanning analysis, and proprietary technical research.
What image resolutions can the Nano banana api generate?
The Nano banana api is highly flexible, supporting resolutions from 512px up to 4K. It also offers specialized aspect ratios like 1:8 and 8:1, which are unique to the 3.1 Flash series. Whether you need a small 0.5K thumbnail or a high-fidelity 4K marketing asset, the Nano banana api adjusts its processing to balance quality and speed. All generated images include a SynthID watermark for responsible AI usage tracking.