Nano Banana 2 Lite
- Model ID
- gemini-3.1-flash-lite-image
- Reference price
- $0.025/ call
- Output sizes
- 1K
- API format
- Gemini native
Quick drafts and frequent 1K work
Image models
Compare the settings available for each model, then open your choice in the Image Studio.
| Model | Model ID | From | Output controls | Editing | Best for | Test route |
|---|---|---|---|---|---|---|
Nano Banana 2 Lite Efficient | gemini-3.1-flash-lite-image | $0.025/ call | 1K | Supported | Quick drafts and frequent 1K work | |
Nano Banana 2 Balanced | gemini-3.1-flash-image | $0.055/ call | 512 / 1K / 2K / 4K | Supported | Everyday generation, editing, and flexible output sizes | Learn more |
Nano Banana Pro Professional | gemini-3-pro-image | $0.09/ call | 1K / 2K / 4K | Not yet verified | Complex instructions and final production assets | Learn more |
GPT Image 2 VIP Exact sizes | gpt-image-2-vip | $0.03/ call | 1K / 2K / 4K | Supported | Pixel-specific output and quality controls | Learn more |
GPT Image 2.5 Flare Exact sizes | gpt-image-2.5-flare-vip | $0.03/ call | 1K / 2K / 4K | Supported | Pixel-specific output and quality controls | |
GPT Image 2.5 Sunburst Exact sizes | gpt-image-2.5-sunburst-vip | $0.03/ call | 1K / 2K / 4K | Supported | Pixel-specific output and quality controls |
Quick drafts and frequent 1K work
Everyday generation, editing, and flexible output sizes
Complex instructions and final production assets
Pixel-specific output and quality controls
Pixel-specific output and quality controls
Pixel-specific output and quality controls
Use the same prompt and reference images in the Image Studio, then choose the result that fits your task and budget.
Test record
This section shows only YingTu test results that are ready for public review. When no result is available, we say so instead of showing draft data.
Whether the product’s shape, proportions, colors, logo, and packaging text remain consistent.
Whether the subject and its edges remain intact while only the requested background changes.
Spelling, capitalization, readability, placement, and any text that was not requested.
Completion, omission, or conflict across instructions that can be checked separately.
The selected dimensions compared with the decoded image size. This does not measure visual quality.
Identity, shape, and color across outputs, plus any unintended changes.