All models
Image generationStable
OpenAI
GPT-Image-2
OpenAI's current image model for detailed generation and editing, especially complex prompts, text-heavy visuals, and high-fidelity source images.
Plain-English overview
What GPT-Image-2 actually is
GPT-Image-2 accepts text and images and produces images through OpenAI's image endpoints. It is aimed at high prompt adherence and detailed compositions rather than a single fixed style.
Cost varies sharply by quality and output size. A low-quality square image can be inexpensive, while a high-quality or 4K output costs many times more. Compare using the setting you will actually ship.
Good fit for
- Text-heavy graphics and infographics
- Product and brand image editing
- High-resolution creative workflows
Category comparison
The facts that matter for image models
These are provider-published specifications, not Cody benchmark scores. Follow the linked sources for current limits and endpoint-specific exceptions.
- Typical price basisA representative current provider price with its quality or resolution basis.
- $0.006–$0.211 at 1024² on fal
- Maximum outputMaximum documented resolution or pixel area for the listed route.
- Up to about 4K / 8.29MP on fal
- Core tasksWhether the model supports generation, editing, or both.
- Generation and editing
- Reference imagesHow the model can use images to guide or edit an output.
- High-fidelity image input
- Text & layoutProvider-documented positioning for typography and structured composition.
- Designed for detailed text-heavy images
- API accessDirect and notable third-party routes included in this guide.
- OpenAI, fal
API and provider access
Where to get GPT-Image-2
Availability
Regions and access stage
Available through OpenAI's Images API and as a partner model on fal.
Direct OpenAI access follows OpenAI's live supported-country list; fal availability follows its own terms.
Check live availabilityData and training
The route matters.
For direct OpenAI API use, API content is not used for training by default and standard retention controls apply. fal stores request input/output by default, but its no-store header can prevent payload storage; uploaded CDN files require separate handling.
This is a concise reading of the cited provider material, not legal advice. A third-party gateway can have different storage, routing, training, and residency terms from the model maker's direct API.
Read the provider policyFrequently asked questions
GPT-Image-2 FAQ
What is GPT-Image-2?
OpenAI's current image model for detailed generation and editing, especially complex prompts, text-heavy visuals, and high-fidelity source images. GPT-Image-2 accepts text and images and produces images through OpenAI's image endpoints. It is aimed at high prompt adherence and detailed compositions rather than a single fixed style.
When was GPT-Image-2 released?
GPT-Image-2 was released on April 21, 2026 according to the cited provider materials.
Where can I access GPT-Image-2?
Available through OpenAI's Images API and as a partner model on fal. The access routes listed in this guide are OpenAI, fal.
How much does GPT-Image-2 cost?
From $0.006 / 1K square image on fal. A 1024×1024 image is listed at $0.006 low, $0.053 medium, or $0.211 high on fal; token use and edits can add cost.
Where is GPT-Image-2 available?
Available through OpenAI's Images API and as a partner model on fal. Direct OpenAI access follows OpenAI's live supported-country list; fal availability follows its own terms.
Is my GPT-Image-2 API data used for training?
For direct OpenAI API use, API content is not used for training by default and standard retention controls apply. fal stores request input/output by default, but its no-store header can prevent payload storage; uploaded CDN files require separate handling. The policy belongs to the provider route and account terms, so verify it again before production use.
Research trail
Sources for this guide
Keep comparing
More image generation models
Google · Stable
Nano Banana 2
Google's versatile image workhorse for fast generation, conversational edits, multiple references, grounded imagery, and outputs up to 4K.
Read guideBlack Forest Labs · Stable
FLUX.2 Pro
Black Forest Labs' production-focused image model for scalable generation and editing, precise composition, and multi-reference creative workflows.
Read guideIdeogram · Stable
Ideogram 4.0
A design-oriented image model with open weights, multilingual typography, structured layout control, editable elements, and simple per-image API tiers.
Read guide