Skip to main content
Nano Banana 2 Lite: the low-cost, high-frequency image route
  • Stable model ID: gemini-3.1-flash-lite-image
  • LaoZhang API price: currently $0.025/call; confirm final pricing and charges in the console
  • Model fit: 1K text-to-image, image editing, batch drafts, real-time interaction, and high-concurrency workflows
  • API contracts: OpenAI-compatible /v1/chat/completions or Gemini-native generateContent

Create an API Key

Create a per-call token and inspect every call record

View the Image API Selection Guide

Compare Standard, Lite, Nano Banana 2, and Pro

Why Lite fits low-cost, high-throughput workloads

Nano Banana 2 Lite is the market name for Google Gemini 3.1 Flash Lite Image. Google’s image generation guide positions it as the image family’s efficiency specialist for low-latency, high-volume interactive applications. Its stable model ID is gemini-3.1-flash-lite-image.
For 2K/4K, video-to-image, or Google Image Search grounding, use Nano Banana 2 API. For the most complex professional assets and precise brand control, use Nano Banana Pro API.

Why the price fits batch workloads

LaoZhang API currently provides a per-call Nano Banana 2 Lite route at $0.025 each. Google’s July 2026 Standard reference is $0.0336 per 1K image. A simple 10,000-call price calculation looks like this:
This is a unit-price multiplication, not a total-cost guarantee. It excludes retries, input tokens, exchange rates, promotions, taxes, and storage. “Cheapest” should never be treated as a permanent claim across every date, region, provider, and contract.

OpenAI-compatible request

Curl

Python SDK

Gemini-native request

The Gemini-native contract uses this route:
Lite only supports 1K. Passing 2K or 4K does not turn it into a high-resolution model; use gemini-3.1-flash-image or gemini-3-pro-image instead.

Image editing request

In the OpenAI-compatible format, place the instruction and reference image URL in the same message:

High-concurrency integration guidance

Lite is designed for high-frequency interaction, but production throughput still requires engineering controls:
  1. Reuse HTTP connections instead of creating one client per image.
  2. Apply jittered exponential backoff to 429s, 5xx responses, and network timeouts.
  3. Use business timeouts and idempotency controls to avoid duplicate billing after uncertain retries.
  4. Log request IDs, model, status code, latency, and console order state.
  5. Before sustained high-volume launch, share peak concurrency, daily calls, prompt size, and input image size with support for capacity confirmation.
The LaoZhang gateway does not publish a fixed low-concurrency plan, but Google upstream can still enforce RPM/IPM, project quota, real-time capacity, and safety policy. This documentation does not promise unconditional permanent unthrottled access.

Global network and CDN note

Applications use the single https://api.laozhang.ai endpoint, while documentation and static assets are CDN-delivered. CDN can improve DNS, TLS setup, and static transfer paths; image generation latency still includes gateway processing, upstream queueing, model inference, and response delivery. Test from your real deployment region.

FAQ

LaoZhang API currently charges $0.025 per call; confirm the price in the console. Google’s July 2026 Standard reference is $0.0336 per 1K image.
Yes. The current stable model ID is gemini-3.1-flash-lite-image, not a preview ID.
No. Lite is 1K-only. Use gemini-3.1-flash-image or gemini-3-pro-image for 2K/4K output.
Yes. It accepts text and image input for local edits, background changes, color replacement, sticker generation, and rapid iteration.
No. Use Nano Banana 2 for Web/Image Search grounding, or Pro for complex professional generation.
Lite targets high-volume use cases, and LaoZhang API does not publish a fixed low gateway concurrency plan. Actual capacity still depends on upstream and account state; confirm capacity before a large launch.
Billing follows the console order state. Review 429s, 5xx errors, timeouts, and validation failures against call logs; a policy-triggered successful response with no image needs separate review.

Nano Banana Standard API

Use the existing Nano Banana API entry and model comparison

Nano Banana 2 API

0.5K–4K, high throughput, and Search grounding

Nano Banana Pro API

Complex composition, professional assets, and 4K

Live Test

Validate prompts and output before integration