ChatGPT Images 2.5: What’s New and How to Use It

ChatGPT Images 2.5 is less about making AI images prettier and more about making the whole image workflow easier to control. Faster generation, stronger reference fidelity, multi-turn editing, and the new Sketch feature make it much more practical for marketers, creators, designers, and everyday ChatGPT users.

Key Takeaways

• OpenAI launched ChatGPT Images 2.5 on September 8, 2026, with sharper detail, more precise editing, and up to 50% lower generation latency versus Images 2.0.
• The biggest workflow addition is Sketch, which lets you draw a rough visual directly in ChatGPT and use it as a guide for generation.
• Images 2.5 is available across ChatGPT, ChatGPT Work, and Codex on desktop, mobile, and web, including all ChatGPT tiers.
• Developers get two API models: GPT-Image-2.5 Flare for faster everyday generation and GPT-Image-2.5 Sunburst for higher-control creative workflows.
• The update is most valuable when you need to iterate on the same asset rather than generate a one-off image and start over.

What Is ChatGPT Images 2.5?

ChatGPT Images 2.5 is OpenAI’s September 2026 update to its image-generation system inside ChatGPT. It is designed to create and edit images from text and reference images while preserving more of the details you care about across revisions. OpenAI says the system now delivers more natural lighting, richer textures, better reference-photo fidelity, and more reliable instruction following.

The practical difference is the editing loop. Instead of treating every prompt as a fresh generation, Images 2.5 is designed to carry context forward. That matters for a product mockup, ad creative, character concept, editorial graphic, or social post where you may need five or ten small changes before the asset is usable.

OpenAI also says more than 3 billion images are created each week across ChatGPT Images and its GPT-Image API models. That scale helps explain why this update is focused heavily on speed and workflow control rather than only on visual quality.

What Changed in ChatGPT Images 2.5?

ChatGPT Images 2.5 improves four areas that are easy to notice in real creative workflows: image fidelity, targeted editing, consistency across multiple edits, and instruction handling. OpenAI also added Sketch, templates, image comments, and prompt sharing.

1. Better reference-photo fidelity

Images 2.5 is built to preserve recognizable subjects more reliably when you use a reference image. OpenAI specifically highlights better retention of distinctive features, more natural lighting and textures, and stronger consistency when a familiar subject is moved into a new setting or style.

For marketers, that makes reference-led creative more useful. You can start with a product photo or existing campaign asset and ask for a new environment without completely rebuilding the visual identity. For creators, it can reduce the number of generations needed to keep a character or subject recognizable.

2. More precise editing

The second major improvement is targeted editing. You can ask Images 2.5 to change a specific element while preserving the surrounding composition. OpenAI gives examples such as changing a product, background, or piece of copy while retaining the rest of the design.

This is where the update feels more like an editing assistant than a basic text-to-image generator. If a product shot is already 90% right, you can focus on the 10% that is wrong instead of prompting for the entire image again.

3. Better multi-turn consistency

OpenAI says Images 2.5 is more reliable during longer conversations. Earlier edits are more likely to remain intact while later instructions are applied. This is a meaningful change because image generation often becomes frustrating when every new edit unexpectedly changes something that was already correct.

In practice, this makes conversational image creation more suitable for production work. You can move from “make the background darker” to “change the headline” to “add more space around the product” without necessarily losing the previous decisions.

4. Stronger layouts and visual instructions

Images 2.5 is also designed to handle more complex visual instructions, real-world information, transparent backgrounds, and specific visual styles. That combination is useful for presentation graphics, product concepts, social creatives, UI concepts, and branded campaign assets.

How to Use ChatGPT Images 2.5

ChatGPT Images 2.5 is available inside ChatGPT on desktop, mobile, and web. You can ask ChatGPT to create an image in a conversation, or use the Images experience from the product interface. OpenAI’s help documentation also confirms that existing images can be uploaded and edited with natural-language instructions.

Step 1: Start with the outcome, not the model

Instead of writing a technical prompt, describe the asset you actually need. For example: “Create a 1200×630 editorial hero image for an article about AI image generators. Use a dark technology aesthetic, a laptop showing a generic image-generation interface, and enough negative space on the left for a headline.”

Step 2: Add a reference when consistency matters

If the visual needs to preserve a product, person, layout, or existing design language, upload a reference image. Then state exactly what must remain unchanged. A useful pattern is: “Keep the subject, pose, camera angle, and overall composition. Change only the background and clothing.”

Step 3: Use Sketch when words are not enough

Sketch is the most distinctive new feature in the ChatGPT Images 2.5 update. You can draw directly in ChatGPT and use the doodle as a visual guide. OpenAI suggests use cases such as sketching a room layout, an outfit contour, or a rough idea that would be difficult to explain precisely in text.

To try it, type @Sketch in ChatGPT. You do not need polished drawing skills. The point is to communicate structure: where the subject goes, where the camera is, what shape the object should have, or how the composition should flow.

Step 4: Edit in small, controlled passes

Once the first version is close, avoid rewriting the whole prompt. Give one or two targeted changes at a time. For example: “Keep everything else unchanged. Replace the blue background with a warm studio gray.” This gives the model a clearer boundary around the requested change.

Step 5: Use comments and share prompts

Images 2.5 also adds comments directly on images for focused edits and lets users share the prompt behind an image. The latter is useful for creative teams because a successful visual can become a reusable starting point rather than a one-off result.

ChatGPT Images 2.5 Pricing and Availability

For ChatGPT users, Images 2.5 is rolling out across all ChatGPT tiers on desktop, mobile, and web. OpenAI also lists availability for ChatGPT Work and Codex. The company does not present a separate consumer “Images 2.5” subscription price; access is part of the relevant ChatGPT product experience and subject to each plan’s usage limits.

Pricing caveat: API pricing is different from ChatGPT subscription access. OpenAI currently lists GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst as API models priced by token usage. Check the live API documentation before budgeting a production workflow because actual image-token consumption can vary with image inputs, quality, and dimensions.

Option	Best for	Current pricing model
ChatGPT Images 2.5	Creators, marketers, everyday image creation and editing	Included within ChatGPT access; plan usage limits apply
GPT-Image-2.5 Flare	Fast, high-quality everyday API generation	$5/1M text input tokens; $8/1M image input tokens; $30/1M image output tokens
GPT-Image-2.5 Sunburst	High-control creative and production workflows	$5/1M text input tokens; $8/1M image input tokens; $30/1M image output tokens

OpenAI describes Flare as the default choice for most API applications and says it delivers higher-quality images than GPT-Image-2 at 50% lower latency. Sunburst is positioned for premium workflows where tighter control is more important than speed.

Where ChatGPT Images 2.5 Actually Helps

The strongest use cases are the ones where an image needs several iterations. A one-shot image generator can already produce impressive results. Images 2.5 becomes more useful when you need to preserve something while changing something else.

Marketing and ad creative

Marketers can use reference-led generation to produce variations of a product campaign while maintaining a recognizable product. This is especially useful for testing different backgrounds, seasonal themes, formats, or visual concepts before moving into final design production.

For a broader overview of image-generation platforms, see AIToolInsight’s AI image tools guide.

Social media content

Templates and faster generation make Images 2.5 practical for recurring social assets. Instead of designing every concept from scratch, you can start with a format, generate the core visual, then iterate conversationally on the details.

If your workflow also involves broader design work, AIToolInsight’s AI visual design tools guide provides useful alternatives and adjacent tools.

Product visualization

Product teams can use image references to explore environments, packaging concepts, campaign scenes, and presentation visuals. The ability to preserve a subject while changing the context is more valuable here than raw artistic style.

Editorial and content production

For publishers, Images 2.5 can generate hero concepts, illustrations, diagrams, and visual variations without requiring a completely separate image workflow. The key advantage is the ability to refine an asset through the same conversation.

Concept development

Sketch makes the tool especially useful during the earliest stage of a project. A rough wireframe or ugly doodle can communicate composition faster than a paragraph of prompt text. That lowers the barrier between an idea in your head and something a team can actually see.

Pros, Cons, and Who It’s Actually For

The biggest strength is control. Images 2.5 is not simply trying to win a beauty contest against other image models. Its more interesting improvement is how it handles revision: preserve the good parts, change the requested part, and keep going.

Pros

• Up to 50% lower generation latency versus Images 2.0, according to OpenAI.
• Better preservation of reference subjects and visual details.
• More precise targeted editing.
• Improved consistency across multiple edits in the same conversation.
• Sketch provides a direct visual-input method when text prompts are limiting.
• Templates reduce the blank-page problem for common creative formats.
• Available across ChatGPT on desktop, mobile, and web.

Cons

• Consumer access still depends on ChatGPT plan limits and availability rather than a separate image-only subscription.
• API budgeting is more technical because image output is token-based and usage varies by generation settings.
• More reliable editing does not mean every complex edit will be perfect; human review is still necessary for production assets.
• OpenAI’s strongest speed and quality claims are company-reported, so they should not be treated as an independent benchmark.

For a marketer, content creator, designer, or small business owner already using ChatGPT, Images 2.5 is an easy upgrade to adopt because it does not require learning a separate creative application. For someone who only needs highly stylized art and rarely edits the same asset, the advantage may be smaller.

How It Compares With Other AI Image Tools

The main difference between ChatGPT Images 2.5 and specialist image platforms is workflow. ChatGPT puts image generation inside a conversational environment, while tools such as Midjourney, Adobe Firefly, and Runway are built around more specialized creative ecosystems.

Tool	Best fit	Key distinction
ChatGPT Images 2.5	Conversational creation and iterative editing	Strong reference workflow, multi-turn edits, Sketch
Midjourney	Art direction and stylized image creation	Strong visual style and creative exploration
Adobe Firefly	Professional design workflows	Deep integration with Adobe’s creative ecosystem
Runway	Multimodal creative production	Strong connection between image and video workflows

FAQ: ChatGPT Images 2.5

Is ChatGPT Images 2.5 free to use?

ChatGPT Images 2.5 is available across ChatGPT tiers, including free access, but usage limits depend on the ChatGPT plan. OpenAI also offers GPT-Image-2.5 models through its API, where developers pay based on token usage. API pricing and consumer usage limits are separate systems.

What is Sketch in ChatGPT Images 2.5?

Sketch is a ChatGPT feature that lets you draw a rough idea directly in the chat and use that drawing as a visual guide for image generation. You can sketch a layout, object, outfit, or composition and then describe the desired style. OpenAI says you can activate it by typing @Sketch.

Is ChatGPT Images 2.5 better for editing existing photos?

Yes. One of Images 2.5’s main improvements is more precise editing of reference images. OpenAI says the model is better at changing a requested element while preserving the subject, composition, background, or other details. It also aims to maintain consistency across multiple edits in the same conversation.

What is GPT-Image-2.5 Flare?

GPT-Image-2.5 Flare is OpenAI’s faster API model for high-quality everyday image generation. OpenAI positions Flare as the default choice for most applications and says it provides the quality, editing, and speed improvements of Images 2.5 with up to 50% lower latency than GPT-Image-2.

What is the difference between GPT-Image-2.5 Flare and Sunburst?

Flare is positioned for fast, high-quality everyday generation, while Sunburst is designed for premium creative workflows that need tighter control over detailed edits. Both accept text and image inputs and produce image outputs, but OpenAI positions Sunburst as the better fit when precision matters more than generation speed.

The Verdict

ChatGPT Images 2.5 is a meaningful upgrade because it tackles the frustrating part of AI image generation: getting from a good first draft to a finished asset without losing what already works.

The new Sketch workflow is particularly useful for people who think visually, while reference fidelity and multi-turn consistency make the tool more credible for repeatable marketing and creative work. If you already use ChatGPT, there is little reason not to try the new image workflow. If you are building an automated image pipeline, compare Flare and Sunburst based on latency, control, and actual token consumption before committing to production volume.