Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)

Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) is an image generation model from Google, listed on api14 under the model id google/gemini-3.1-flash-lite-image. It sits in the image category with text as a secondary category, and Google positions it as its fastest, most cost-efficient Gemini image model, aimed at high-velocity developer pipelines and rapid-fire visual exploration. The provider release date is 2026-06-30, and the model's knowledge cutoff is 2025-01-01.

The model accepts image and text as inputs and can return image and text as outputs, so it handles both text-to-image generation and prompts that include reference images. It supports up to 14 input references, which lets you steer a generation with example images rather than text alone. The context window is 65,536 tokens and the maximum output is 58,982 tokens. Generated images are produced at 1K resolution, and you can request one of fourteen aspect ratios: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9 and 21:9. The n parameter accepts a single image per request (minimum 1, maximum 1).

Those capabilities point to a few practical uses. Rapid visual exploration and concept iteration fit the model's stated focus, since you can generate a single 1K image per call across many aspect ratios, from square social formats to wide 21:9 or tall 9:16 layouts. Reference-driven generation with up to 14 input images suits tasks where you want an output that follows the style, subject or composition of supplied examples. Because the model also returns text, it can be used in mixed workflows where a prompt produces both a visual and a textual response. The 65,536 token context window gives room for longer prompts or several reference images alongside instructions.

You call the model through api14's OpenAI-compatible API using the model id google/gemini-3.1-flash-lite-image. The supported request parameters are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, seed, temperature and top_p. Image-specific controls cover resolution (1K), aspect_ratio and n, plus input_references for supplying up to 14 reference images. Because the interface is OpenAI-compatible, existing OpenAI client libraries and request shapes work with this model id.

The main constraints to keep in mind are the single image per request (n is fixed at 1) and the fixed 1K resolution, so work that needs multiple images in one call or higher resolutions is not covered by this model. The 65,536 token context window is also modest compared with text-focused models, which matters if you plan to pass many large reference images or very long prompts. For those cases, another model in the catalogue may fit better.