Image models
The image node lets you choose an image generation model for the job. Compare them for the result you want, such as quick drafts, precise editing, or designs that include text.
Models at a glance
Each model has different strengths. Use the table below to quickly find the model that best fits your task.
| Model | Specialty | Recommended for | Choose it when |
|---|---|---|---|
| GPT Image 2.5 | High-quality generation · precise editing · image consistency | Generation from reference images, repeated edits, keeping people and products, brand visuals and image series | You need several rounds of edits that keep the key elements of a reference image, with fast and consistent results |
| GPT Image 2 | High-quality general generation · editing | Ad creative, product visuals, complex compositions, high-quality images | You need fast generation, high quality, and strong instruction following all at once |
| Nano Banana 2 | Balance of general performance, speed, and efficiency | Quick drafts, repeated generation, everyday image work | Quality, generation speed, and cost efficiency all matter |
| Nano Banana Pro | Professional design · text · information graphics | Posters, infographics, product mockups, images containing text | Complex instructions, readable text, and professional design results matter |
GPT Image 2.5
An image generation and editing model that combines sharp detail and natural rendering with fast generation. It edits precisely while preserving the characteristics of a reference image, so it suits work that follows complex instructions and keeps results consistent across several rounds of editing.
Built for refined generation and repeated editing
- Sharp detail with natural lighting and rich texture
- Keeps key features of a reference image consistent, such as people, products, and places
- Changes only the part you want while preserving the rest of the composition and detail
- Holds earlier changes and image quality steady across repeated edits
- Supports varied image setups, including transparent backgrounds and complex layouts
Flare and Sunburst: In the image node, GPT Image 2.5 comes in two versions, Flare and Sunburst. They cost the same credits and share the same quality levels; Flare generates faster and is selected by default.
GPT Image 2
A general-purpose image generation and editing model that combines fast speed with high quality. It suits work that has to follow complex instructions and produce high-quality results.
Built for high-quality image production
- High-quality image generation and editing
- Follows complex prompts and instructions
- Production-ready images in a range of sizes and ratios
- Image work that uses the detail in a reference image
Nano Banana 2
A general-purpose model that balances image quality, speed, and cost efficiency. It works well for producing several drafts quickly or generating images repeatedly.
Built for fast, general image work
- General image generation and editing
- Balance between speed and result quality
- Suited to repeated generation and quick drafts
Nano Banana Pro
A model built for professional image production and complex instructions. It suits work where layout and clear communication matter, such as designs with text in the image, infographics, and product mockups.
Built for professional design and text
- Complex graphic design and professional assets
- Accurate, readable text inside the image
- Infographics and data or information graphics
- High-quality product mockups and complex production instructions
Which model should I choose?
Pick the model from the item that comes closest to the work you want to do.
I want high-quality images across the board.
High-quality general image production, including complex requests
GPT Image 2I want to edit an existing image while keeping the person or product.
Precise edits to the parts you need while preserving the original’s key elements
GPT Image 2.5I want to try out several images quickly.
Repeated generation that balances quality, speed, and efficiency
Nano Banana 2I want images where text matters, like posters or infographics.
Complex design, text rendering, and information graphics
Nano Banana ProChoosing a model in the image node
The actual image generation happens in the image node in Canvas. You enter a prompt, set the model and aspect ratio, and generate the image. Quality can be set only with GPT Image models.
What the image node and AI Models each cover
The image node page explains how to select a model and generate images in practice, while AI Models · image models compares what each model is good at and which tasks it suits.
Frequently asked questions
Should I always pick the newest model?
No. The right model depends on the task. Choose based on what you need — fast repeated generation, precise editing, designs with text, and so on.
Which model suits a poster with text in it?
When text rendering and complex graphic layout matter, consider Nano Banana Pro first.
Which model should I use to edit an uploaded image?
When you connect the uploaded image to another image node, that node switches to edit mode, and every image model you can currently select can be used for editing. GPT Image 2.5 suits precise edits that keep the key elements of the existing image.
I want to create several drafts quickly.
If you want a balance of general performance, speed, and efficiency, Nano Banana 2 is a good fit.