Skip to main content
nano-banana is Google’s lightweight image generation model exposed through the same OpenAI Images protocol surface as the rest of the gateway’s image tier. The trade-off relative to gpt-image-2: materially cheaper per image, with somewhat looser fidelity on in-image typography and complex composition. The right default when you’re generating at volume — programmatic thumbnails, A/B test images, internal mockups. Pricing: $0.018 / image — see the rate card. Failures don’t bill; pricing sits below gpt-image-2 and above the lowest seedream tier.

​
Protocols

ProtocolPath
OpenAI ImagesPOST https://llm.bytespike.ai/v1/images/generations

​
Quickstart

curl https://llm.bytespike.ai/v1/images/generations \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "nano-banana",
    "prompt": "Minimalist line drawing of a cat sitting in a window, morning light",
    "n": 4,
    "size": "1024x1024"
  }'

​
Capabilities

CapabilitySupported
Text-to-image✅
n ≥ 2 batch generation✅
size 1024×1024 / 1024×1792 / 1792×1024✅
Image-to-image— (use nano-banana-v2)
response_format url / b64_json✅
Modalityimage
Capability bucketimage_generate

​
When to use

  • Volume work — n=4 batches for A/B testing, thumbnail generation, programmatic asset pipelines.
  • Cost-sensitive defaults — when an internal tool needs an illustration and the budget can’t carry gpt-image-2.
  • Acceptable for typography-light scenes — landscapes, abstract, minimalist illustrations.
When not to use:

​
Next