Which image model should I use?

No one model is right for every picture. They differ in what they work from, the sizes and shapes they make, and what a picture costs, and the model menu on Images shows all three for each one before you pick it. Hover over a model in the menu to read what it's for.

Models can be switched off for a while, so your menu may list fewer than this page does. What's in the menu is what you can use now.

Pick by what you need

• A picture from words. The bar starts on Nano Banana, the default. Any model under Make a picture starts from words.
• A person, a product or a place kept from your pictures. Pick a model with image refs: up to 16 on GPT Image 2, GPT Image 2.5 and the three GPT Image 2.6 models, up to 10 on Nano Banana 2, Nano Banana 2.1, FLUX 3 Klein and both Seedream 5.0 models, and up to 8 on Nano Banana Pro and both FLUX 2 models. See Use reference pictures and your own uploads.
• Words in the picture. The menu describes Nano Banana Pro as rendering text better, Nano Banana 2.1 as making it clearer, and Seedream 5.0 Pro as made for readable text and infographics.
• A long, detailed prompt. GPT Image 2.5 is described as faithful to long prompts.
• A poster, a product shot or anything laid out. GPT Image 2.6 is described as having stronger layout. GPT Image 2.6 Sunburst is its refined version, steadier when you edit a picture again and again, and GPT Image 2.6 Economy is the same model at a low flat price, at 1K only.
• Several takes to choose from. Midjourney, Midjourney V7 and Midjourney Niji 7 make four pictures from every prompt, for one price, at Fast or Turbo. Niji 7 is made for anime and illustration. See Make pictures with Midjourney.
• A big picture. Nano Banana Pro, Nano Banana 2, Nano Banana 2.1, GPT Image 2, GPT Image 2.5, GPT Image 2.6, GPT Image 2.6 Sunburst and FLUX 3 Klein make 4K. To go bigger, or to enlarge any picture, upscale it in Edit & Enhance: see Upscale, cut out, relight and edit a picture.
• A shape other than 16:9, 9:16 or 1:1. Nano Banana, Nano Banana Pro, Nano Banana 2, Nano Banana 2.1, GPT Image 2, GPT Image 2.5, the three GPT Image 2.6 models, FLUX 2 Pro and Flex, FLUX 3 Klein, both Seedream 5.0 models, the three Midjourney models and Nano Banana Edit offer every shape the bar has, from Story (9:16) to Classic (4:3) and Photo (3:2). On a Creative Board, a picture card offers each model's other shapes too, such as 21:9, GPT Image 2's 2:1 and 8:1 on Nano Banana 2 and 2.1. Output lists only the shapes the model you've picked makes.
• A change to a picture you already have. Open the Edit a picture tab. The models made to edit come first and need the picture to start from: Nano Banana Edit takes it plus up to 9 more, and Qwen Image Edit takes it alone. The models that make from words and pictures, such as Nano Banana 2.1 and GPT Image 2, can edit it too.

The models

How the menu describes each one, and what it takes:

What they cost

Each row in the menu shows its price: one figure where every size costs the same, and "from" the lowest where the sizes differ. Pictures start from 1 credit, and a Nano Banana picture is 2 credits. A bigger size usually costs more on the same model, and Output prices each one. Seedream 5.0 Pro also adds a little for each picture you give it past the first, and GPT Image 2.6 and GPT Image 2.6 Sunburst about 1 credit for each picture you give them. GPT Image 2.6 Economy costs the same with or without pictures. A Midjourney price is for the run's four pictures together, and Turbo costs more than Fast.

The price on Generate is the one you pay, for the model, size and pictures you've chosen. Models lists every model the site offers.