ChatGPT Images / GPT Image: Features, Pricing, Models & Alternatives (2026)

Last researched: September 26, 2026. OpenAI's own pricing and product pages (openai.com) returned an HTTP 403 to automated fetch during this research; figures here are cross-referenced from independent trackers and should be verified at OpenAI's own pricing page before making purchasing decisions. OpenAI's image generation is exposed to end users as ChatGPT Images, built directly into ChatGPT's chat interface, and to developers as the GPT Image model family available through OpenAI's API.

What Is ChatGPT Images / GPT Image?

ChatGPT Images is the name for OpenAI's built-in image generation and editing experience inside ChatGPT, powered by the GPT Image model family. It replaced ChatGPT's earlier DALL-E-based image generation starting in March 2025, bringing far more accurate in-image text rendering and better prompt adherence. The same underlying models are exposed separately to developers through OpenAI's API under model IDs such as gpt-image-2, letting other applications call the same image-generation and editing capability directly.

Key Capabilities

• Native, comparatively accurate text rendering inside generated images
• Conversational, multi-turn image editing directly within a ChatGPT chat (describe a change in plain language and get an updated image)
• Precise mask-based inpainting and outpainting through the API for developers who need pixel-level control
• Object removal, background replacement, and style transfer via natural-language instructions
• A Sketch feature (added with ChatGPT Images 2.5 on September 8, 2026) that turns a user's rough drawing into a finished AI image
• Architecture that generates images more like a language model generates text, rather than the traditional diffusion approach used by earlier DALL-E models

Main Features

Use Cases

• Generating marketing graphics, illustrations, and social media images from within a chat workflow
• Iteratively editing an existing image through natural-language follow-up instructions
• Turning a rough sketch into a polished image
• Developers embedding image generation/editing into their own apps via the API
• Precise, localized image edits (object removal, background swaps) using mask-based inpainting

Supported Models / Technologies

GPT Image 2 is OpenAI's current flagship image model as of this writing, using an architecture OpenAI and third-party analysts describe as generating images autoregressively (more like an LLM generating text) rather than through the diffusion process used by DALL-E 2/3. Reported native output is up to 2048x2048 (2K), with 4K available in beta per developer trackers; this could not be independently confirmed on OpenAI's own documentation because openai.com blocked automated fetch. GPT Image 1 remains available until its October 23, 2026 shutdown. DALL-E 2 and DALL-E 3 have been fully removed from the API as of May 12, 2026 and are legacy/historical at this point.

Platforms

• Web (chat.openai.com)
• iOS app (ChatGPT)
• Android app (ChatGPT)
• API (developer access to GPT Image models)

OpenAI does not, as of this writing, ship a separate dedicated desktop application solely for image generation; access is through the general-purpose ChatGPT apps or the API.

Integrations

GPT Image models are consumed by many third-party products through OpenAI's API, but OpenAI does not publish an official, exhaustive list of first-party integration partners for image generation specifically. The primary supported integration surface is the OpenAI API itself, which developers can wire into their own applications, plus the built-in experience inside ChatGPT.

Pricing

OpenAI's own pricing pages blocked automated fetch during this research; the figures below are cross-referenced from independent trackers as of September 26, 2026 — verify current figures at OpenAI's official pricing page (openai.com/pricing / platform pricing documentation) before budgeting.

Advantages

• Tight integration into the widely used ChatGPT conversational workflow, so image generation sits alongside text, code, and other tasks
• Comparatively strong native text rendering inside generated images
• Precise, mask-based editing tools for developers who need pixel-level control
• Large existing developer ecosystem already built around the OpenAI API
• Mobile apps (iOS and Android) in addition to the web interface

Limitations

• No standalone desktop app dedicated to image generation
• OpenAI's own pricing and marketing pages blocked automated fetch during this research, meaning some figures here rely on third-party cost trackers rather than the primary source
• Legacy models (DALL-E 2/3, and soon GPT Image 1) are being actively sunset, which can require migration work for existing integrations
• Content policy restrictions apply (no public figures without permission, no depictions of copyrighted characters, no explicit content)
• Per-image API cost varies by quality tier and reported figures differ somewhat between independent trackers, making exact budgeting harder without checking OpenAI's own current documentation

Security & Privacy

For consumer ChatGPT accounts, OpenAI states content may be used to improve its models unless the user opts out via Settings > Data Controls > "Improve the model for everyone." For API usage, OpenAI states customer-submitted data is not used to train or improve its models unless the customer explicitly opts in. ChatGPT Enterprise, Team, and API Platform accounts carry contractual data processing agreements that prohibit using submitted data for training. Generated content is subject to OpenAI's usage policies and automated content moderation; consult OpenAI's own Terms of Use and privacy documentation for full specifics, since some detail could not be verified directly against openai.com during this research.

Best For

Users and developers already working inside the ChatGPT or broader OpenAI ecosystem who want image generation and iterative, natural-language editing available in the same place as their text and code workflows, or who need an API-accessible image model with strong prompt adherence and in-image text rendering.

Alternatives

• Midjourney — a dedicated image generator with a distinctive visual style, accessed via Discord or its own web app
• Adobe Firefly — emphasizes commercially-safe, licensed training data and deep integration into Creative Cloud apps
• Google Gemini Image Generation (Nano Banana) — strong multimodal understanding and conversational multi-turn editing
• Ideogram — known for particularly strong in-image text rendering
• Stability AI (Stable Diffusion / Flux) — open-weight models that can be self-hosted

Frequently Asked Questions

What is GPT Image 2?

GPT Image 2 is OpenAI's flagship image generation and editing model, released in April 2026 as the successor to GPT Image 1 and, before that, DALL-E 3. Its model ID for API use is gpt-image-2.

Is DALL-E still available?

DALL-E 2 and DALL-E 3 were removed from OpenAI's API on May 12, 2026. ChatGPT itself moved from DALL-E to the GPT Image model family starting in March 2025. GPT Image 1 is scheduled for shutdown on October 23, 2026, with GPT Image 2 as its replacement.

Can I opt my data out of being used for model training?

Yes, for consumer ChatGPT accounts: go to Settings > Data Controls and turn off 'Improve the model for everyone.' For the API, OpenAI states it does not use customer-submitted data to train models unless the customer explicitly opts in. ChatGPT Enterprise, Team, and API Platform accounts have contractual data processing agreements that prohibit training use.

Does OpenAI offer a free image generation tier?

ChatGPT's Free tier includes limited image generation through GPT Image, subject to usage caps. Full access sits behind the Plus ($20/month) and Pro ($200/month) ChatGPT subscriptions, while developers pay separately per image through the API.

What editing capabilities does GPT Image 2 support?

GPT Image 2's editing endpoint supports precise inpainting and outpainting through mask images, letting a specific region of an image be changed while the rest stays untouched, along with object removal, background replacement, and style transfer described in plain language.

Final Summary

OpenAI's image generation, presented to consumers as ChatGPT Images and to developers as the GPT Image model family (currently GPT Image 2), is defined by its tight integration into ChatGPT's conversational workflow and its API-accessible, mask-based editing tools, with older DALL-E models being actively retired in favor of the GPT Image line.