GPT-IMG
It follows the awkward parts of a prompt
Object counts, spatial relationships, and "everything except this one thing" clauses survive further into a long brief than they used to.
GPT-IMGEnglish
GPT Image 2.5 generator
The newest OpenAI image model, ready to run. Tighter instruction following, cleaner text inside the frame, and edits that keep your subject recognizable.
Run GPT Image 2.5 in your browser. Write a prompt or upload a photo to edit, choose 1K to 4K, and see the credit cost before every render.
OpenAI GPT Image 2.5Prompt or reference inputUp to 4K outputThree quality tiers
149 / 20,000
Image
GPT Image 2.5 next to the model it replaces
Version 2.5 is the default here. GPT Image 2 stays on the menu for cheaper bulk passes.
What GPT Image 2.5 actually changes
The 2.5 release targets the two places image models usually fall apart: complicated instructions, and text inside the picture.
GPT-IMG
Text in the picture stays readable
Signage, packaging copy, menu boards, and book covers hold their letterforms instead of collapsing into decorative squiggles.
GPT-IMG
Edits stay scoped to what you asked for
Change one element and the rest of the frame is left alone, which is what makes several rounds of revision practical instead of a lottery.
GPT-IMG
Busy layouts hold together
Multi-panel compositions, product line-ups, and crowded scenes keep their internal logic rather than drifting apart at the edges.
GPT-IMG
From text alone
Type the brief and generate. Best when nothing exists yet and you want the model to make every compositional decision.
GPT-IMG
From your own image
Attach a reference and describe the change. The request routes to the editing endpoint, so the original composition is the starting point rather than a suggestion.
Keep exploring GPT-IMG
One account, the full image generation workflow.
Choose Your Plan
Upgrade for larger monthly credits, faster generation, and commercial-ready output.
Need more without a subscription?
Buy credits once and use them whenever you need. Purchased credits never expire.
FAQs
Practical details for this generator page.
It is the latest generation of the OpenAI image model, covering both text-to-image and editing. GPT-IMG runs it directly, so you do not need your own API key.
Version 2.5 holds up better on long, fussy prompts, renders text inside the image more reliably, and keeps edits confined to the part you named. Version 2 remains the cheaper option for simple, high-volume work.
It is priced identically to GPT Image 2: 10 credits for a 1K standard render, up to 120 for 4K at high quality. Editing an uploaded image starts at 15 credits. The exact figure is shown before you generate.
Yes. Upload up to five reference images, describe the change, and the job goes to the editing endpoint instead of generating from nothing.
Up to 4K, across all seven aspect ratios. Larger sizes cost more credits and take longer, so drafting at 1K and finishing at 4K is usually the cheaper habit.
Start with GPT Image 2.5 generator
Use the prompt panel above with the same account, credits, and history workflow.