ChatGPT Images 2.5: What’s New, API Models, Pricing & How to Use It

01

Overview

OpenAI launched ChatGPT Images 2.5 on September 8, 2026, upgrading both the image model and the way people create and edit images inside ChatGPT. The release combines model-level improvements—sharper detail, more reliable edits, stronger subject preservation and faster generation—with new product tools such as Sketch, templates, image comments and prompt sharing. OpenAI also released two developer models: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst.

02

What is ChatGPT Images 2.5?

ChatGPT Images 2.5 is OpenAI’s latest image-generation and image-editing model for ChatGPT. OpenAI describes it as its new state-of-the-art image model and says it improves natural lighting, texture quality, reference-photo fidelity, complex instruction following and editing consistency across multiple turns.

The most important practical change is not simply that generated images look better. Images 2.5 is designed to preserve more of an existing image when you request a focused edit. That matters for creators, marketers and designers who need to change one product, background, label, piece of copy or visual detail without rebuilding the entire asset.

03

What changed in Images 2.5?

OpenAI highlights five main areas: better reference-image fidelity, more precise editing, stronger consistency over several edits, improved handling of complex layouts and visual instructions, and faster generation. OpenAI says generation latency is up to 50% lower than Images 2.0. That figure is a vendor claim and should not be treated as a guaranteed speedup for every prompt or device.

Images 2.5 also handles transparent backgrounds and complex layouts more reliably, according to OpenAI. For production workflows, this can make it more useful for product imagery, social graphics, campaign assets, UI concepts, presentation visuals and repeated brand-oriented edits.

04

New ChatGPT image tools: Sketch, templates, comments and prompt sharing

The model release is paired with several ChatGPT product features. Sketch lets you draw a rough visual guide directly in ChatGPT and turn it into a generated image. OpenAI says you can launch it by typing @Sketch where the feature is available.

Templates provide starting points for common creative formats such as posters, merchandise and product imagery. Image comments let you attach focused editing instructions to an image, while prompt sharing lets another user reuse the prompt behind a generated image and adapt it with their own details.

These are product features around the model, not separate image models.

05

Availability

OpenAI says Images 2.5 is rolling out across all ChatGPT tiers on desktop, mobile and web. It is also available in ChatGPT Work and Codex. Individual accounts can still experience staged rollout behavior, so a user may not see every interface feature at exactly the same moment.

06

GPT-Image-2.5 Flare vs Sunburst

OpenAI released two API models. GPT-Image-2.5 Flare is positioned as the faster default for everyday high-quality generation and editing. OpenAI says it brings the 2.5 quality and editing improvements with 50% lower latency than GPT-Image-2 and recommends it for creator content, social content, product experiences, visual search and high-volume generation.

GPT-Image-2.5 Sunburst is positioned as the more precise option for premium creative and editing workflows. OpenAI recommends it when tighter control across edits matters more than speed, such as polished campaign creative or product imagery.

That does not mean Sunburst is automatically the better choice for every task. The useful question is whether extra precision is worth the longer generation time for a specific workflow.

07

API pricing

OpenAI’s current model cards list the same token rates for Flare and Sunburst: text input $5 per 1 million tokens, cached text input $1.25, image input $8, cached image input $2, and image output $30 per 1 million tokens. Text output is not billed because these models output images.

OpenAI does not publish one universal flat cost per finished image. Actual spend depends on token usage, resolution, quality setting, references and the number of retries or edits. Any article claiming a single fixed official cost per image should be treated cautiously unless it specifies the exact configuration and source.

08

Quality settings and API use

The OpenAI model pages list low, medium, high, xhigh, max and auto quality settings for both models. Developers can select the models in the Images API or use them through the Responses API image-generation tool.

For production systems, one reasonable pattern is to use Flare for exploration and higher-volume generation, then reserve Sunburst for assets where edit precision or final polish matters. That workflow is an inference from the documented positioning, not a rule imposed by OpenAI.

09

Creator and marketing use cases

Images 2.5 is particularly relevant for social content, product mockups, campaign concepts, creator thumbnails, branded asset variations, presentation visuals, transparent-background assets and iterative edits where preserving a subject or composition matters.

The stronger multi-turn editing behavior may be more important than one-shot generation quality for real marketing work. A production asset often goes through several revisions, and repeated full regeneration can introduce drift in faces, products, typography, layout or brand treatment.

10

Limitations and verification notes

OpenAI’s performance statements are first-party claims. Independent launch-day testing exists, but there is not yet a large body of standardized public benchmarks covering every style, language, editing task and resolution.

The up-to-50% latency improvement is not a universal wall-clock guarantee. ChatGPT plan limits can also vary and OpenAI has not published one universal numeric quota for every tier. API cost should be calculated from token usage rather than a made-up per-image figure.

11

FAQ

Is ChatGPT Images 2.5 free?

OpenAI says Images 2.5 is available across all ChatGPT tiers, including Free. Usage limits still depend on the account and plan.

What is ChatGPT Sketch?

Sketch is a ChatGPT feature that lets you draw a rough idea and use it as a visual reference for the generated result.

What are GPT-Image-2.5 Flare and Sunburst?

They are the two API models released with Images 2.5. Flare is positioned for faster everyday use; Sunburst is positioned for higher-precision creative workflows.

Does Images 2.5 support editing existing images?

Yes. OpenAI specifically emphasizes focused edits, reference-image fidelity and better consistency across repeated editing turns.

How much does the API cost?

Both current OpenAI model cards list $5 per 1M text-input tokens, $8 per 1M image-input tokens and $30 per 1M image-output tokens, with cached-input discounts. There is no single universal official per-image price.

Is Images 2.5 available in Codex?

Yes. OpenAI says the rollout includes Codex as well as ChatGPT and ChatGPT Work.

12

Sources

Primary facts should be rechecked against OpenAI’s launch post, Help Center and live API model cards immediately before publication.