Image model · by OpenAI
Avocado's default image model. Create product images from a prompt or make precise edits to your own product photo with OpenAI's GPT Image 2.5 Flare and Sunburst.
GPT Image 2.5 is OpenAI's image generation and editing model, released in September 2026 in two versions, Flare and Sunburst. On Avocado AI, GPT Image 2.5 Flare is the default image model for product images and edits, with 1K, 2K, 4K and custom sizes.
For ecommerce brands
Campaign images of your real product, built from your own photo and references.
Change the scene, props or light while the label and packaging stay recognisable.
UGC starts with a GPT Image 2.5 edit that puts your real product in the creator's hand.
Specs on Avocado
| Made by | OpenAI |
|---|---|
| Versions on Avocado | Flare (default) and Sunburst |
| Modes on Avocado | Text to image, image edit |
| Sizes | 1K, 2K, 4K and custom sizes |
| Quality levels | Low, medium, high, X-high, max |
| Reference images per edit | Up to 14 |
| Credits per 1K image | 1 at low, 2 at medium, 4 at high quality |
| Plan access | Every paid plan; 4K and max quality from Starter |
Credits are the live workbench prices. The Generate button always shows the exact cost before you create. Plans in EUR, see pricing.
GPT Image 2.5 Flare handles product edits well: it follows detailed instructions and holds on to the product in your reference photo. That is why Avocado uses it by default in the image workbench, in edits made by Super Agent and for putting your product in a creator's hand in UGC.
OpenAI released GPT Image 2.5 as two API models. Flare is the general default for most work. Sunburst is aimed at premium creative and editing work where you want tighter control across edits. Both are in the Avocado image picker and share the same credit table.
At 1K, a GPT Image 2.5 image costs 1 credit at low quality, 2 at medium and 4 at high. 2K and 4K sizes and the max quality level cost more, and the Generate button shows the exact price before you create. Plans start at EUR 19.99 a month for 100 credits.
Upload the product photo, choose GPT Image 2.5 Flare Edit and describe the change. Add up to 14 references, such as a label close-up or a mood image. Set the quality, generate, and compare the versions with your original. Keep the best one or edit it again.
Updated September 26, 2026
Yes. Choose 1K, 2K, 4K or a custom size in the settings. 4K and the max quality level need a Starter, Growth or Pro plan. Larger sizes cost more credits, and the settings dialog shows the price of each option before you generate, so there are no surprises.
They come from the same OpenAI release. ChatGPT Images 2.5 is the experience inside ChatGPT, and GPT Image 2.5 Flare and Sunburst are the API models behind it. Avocado uses the API models, so you get them next to Nano Banana 2, Seedream and every video model.
OpenAI built GPT Image 2.5 to better preserve the people and products in reference photos, which is why it is Avocado's default for edits. Results still vary with the prompt, so add a label close-up as a reference, generate a few versions and compare them with the original.
Yes. Every paid plan includes full commercial rights and no watermark, so you can run what you make in paid ads, on product pages, in email and on social. Results are saved to your Avocado library, ready to reuse in a video, a Flow or a Super Agent post.
Yes. When Super Agent edits your product photos or builds a creator look for a UGC video, it uses GPT Image 2.5 Flare by default. You can also ask it for a different image model in the chat, such as Nano Banana 2 or Seedream, for a specific task.
Yes, through the Avocado MCP server. Connect it to Claude, Codex, Cursor or your own agent and ask for an image or an edit. GPT Image 2.5 Flare is the default model for the edit tool, it runs on your Avocado credits, and results land in your library.
More models
Google's fast image model for product shots and edits, up to 4K.
Google DeepMind's video model with native audio and start and end frame control.
ByteDance's video model for product ads and UGC, 4 to 30 seconds with native audio.
Kuaishou's video model in Standard, Pro and native 4K, 3 to 15 seconds.
GPT Image 2.5 and every other image and video model in one workspace, on one credit balance.
