OpenAI's flagship natively multimodal image model with industry-leading instruction following, precise region-aware editing, and best-in-class text rendering — now up to 4x faster than its predecessor.
Built on a unified transformer backbone that processes text and image tokens natively — not a diffusion model — enabling superior reasoning and instruction following.
Modify specific parts of an image while preserving faces, logos, lighting, and composition exactly as they are. Accepts up to 16 input images per request.
Generates legible, correctly styled text at small point sizes with multi-line support up to 800 characters — ideal for posters, banners, and branded graphics.
GPT Image 1.5 is OpenAI's flagship image generation model and the successor to GPT Image 1, released in December 2025. Unlike traditional diffusion-based models such as DALL-E 3 or Stable Diffusion, GPT Image 1 and 1.5 use a natively multimodal autoregressive architecture — the same transformer backbone processes both text and image tokens together. This means the model genuinely reasons over prompts rather than simply conditioning a diffusion process, which translates into dramatically better instruction adherence, spatial composition, and layout control. The 1.5 version brings generation speeds up to 4x faster than GPT Image 1, costs approximately 20% less per API call, and introduces region-aware editing that can surgically alter one element while keeping everything else pixel-perfect.
Instruction following is where GPT Image 1.5 truly shines. The model can handle intricate, multi-step prompts — such as "create a 6×6 grid of specific icons and symbols" — and follow them accurately, a task where most competing models fail. Text rendering has been substantially improved over both GPT Image 1 and earlier generation models: the model supports dense, small-point-size text with correct font weight and style, making it suitable for newspaper layouts, poster typography, and UI screenshots. Facial and logo consistency across iterative edits is another standout: when you modify one element of an image, the model preserves lighting, composition, and likeness in the untouched areas — addressing the common "slot machine" problem where older models would regenerate everything with every edit.
| Feature | GPT Image 1 | GPT Image 1.5 |
|---|---|---|
| Architecture | Autoregressive multimodal | Autoregressive multimodal |
| Generation speed | ~30–60 seconds | 10–30 seconds (up to 4x faster) |
| API pricing | Baseline | ~20% cheaper |
| Text rendering | Strong | Improved — denser, smaller text |
| Editing precision | Good | Region-aware, element-specific |
| Max input images | 16 | 16 |
| Output resolutions | 1024x1024, 1024x1536, 1536x1024 | 1024x1024, 1024x1536, 1536x1024 |
| Quality tiers | Low / Medium / High | Low / Medium / High |
| Transparent backgrounds | Yes (PNG) | Yes (PNG) |
| C2PA provenance metadata | Yes | Yes |
Both variants available here — GPT Image 1.5 (text-to-image) and GPT Image 1.5 I2I (image-to-image) — are powered by the 1.5 model. Use text-to-image for new creations and I2I for editing or style-transferring an existing image.
| Variant | credits |
|---|---|
| GPT Image 1.5 | 10 |
| GPT Image 1.5 I2I | 10 |
1 credit = $0.012
Generate consistent on-brand graphics, ad creatives, and product visuals at scale with accurate logo and color preservation across edits.
Produce product image variants — different scenes, angles, and backgrounds — from a single source image without reshoots.
Create posters, banners, UI mockups, and infographics where readable, correctly styled text is embedded directly in the image.
PopularBlack Forest Labs
Black Forest Labs' production-grade image generation model family delivering 4MP photorealistic output, multi-reference consistency across up to 10 images, and reliable text rendering — all in sub-10-second generation speeds.
From 3 credits
FastGoogle's Gemini Flash-powered image generation and editing model that went viral for its speed, real-world knowledge, and AI-assisted editing capabilities.
From 2 credits
Google DeepMind's leading text-to-image model delivering up to 2K resolution, superior text rendering, and diverse art styles — engineered for professional creative work.
From 2 credits
Start generating amazing content with GPT Image 1.5 today
Try GPT Image 1.5 Now