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

Flare is the default API choice: higher quality than GPT-Image-2 at 50% lower latency — external evaluation by Manus measured it delivering high-quality images 2 to 4 times faster than GPT-Image-2. Sunburst adds an extra level of precision for detailed creative work like production ad creative and polished product images, at the cost of longer generation times. Both are available in the API now.

How much does GPT Image 2.5 cost?

OpenAI's launch announcement links to its pricing page but does not state per-image or per-token prices. ChatGPT image generation is included in existing plans across all tiers — Free, Plus, Pro, Business, and Enterprise. For API pricing, check OpenAI's official image generation pricing page before committing to a budget, as rates for the two new variants have not been announced in the launch post.

What are the Sketch and Templates features?

Sketch is a new ChatGPT feature — type @Sketch to draw directly in the chat interface, and your doodle becomes a visual reference the model follows when generating the final image. Templates provide starting points for popular formats like posters, merch, flyers, and product photos so you don't start from a blank canvas. You can also place comments directly on images to focus edits, and share images with their prompts attached.

How does GPT Image 2.5 compare to Midjourney?

They optimize for different workflows. GPT Image 2.5 excels at conversational, iterative editing — multi-turn changes that preserve earlier edits — and integrates with ChatGPT's broader assistant context. Midjourney remains strong on stylized, aesthetic-first generation and its community tooling. If your work is 'generate, then refine over many turns,' GPT Image 2.5 fits; if it's single-shot artistic output, Midjourney is still worth testing side by side.

GPT Image 2.5 Review: OpenAI's Faster, Sharper Image Model (2026)

What is GPT Image 2.5?

GPT Image 2.5 is OpenAI's newest image generation model, announced September 8, 2026 and rolling out the same day to every ChatGPT surface — Free, Plus, Pro, Business, and Enterprise on desktop, mobile, and web — plus two new API variants. It arrives nearly a year and a half after the original GPT-4o image generation moment, into a market where OpenAI says users now create more than 3 billion images per week across ChatGPT and the API.

The headline change is speed: up to 50% lower generation latency than Images 2.0. But the more meaningful upgrades are in editing behavior — and that's where this model is aimed.

Key Features

Editing That Doesn't Break Things

The core complaint about AI image editing has always been collateral damage: ask for one change, lose three others. Images 2.5 targets exactly this:

Edits only what you asked for, keeping the rest of the image intact — even with complex subjects and backgrounds
Follows editing instructions more reliably across a sequence of changes
Multi-edit consistency: early changes hold up over long sessions "without degrading image quality over time"

For anyone who has tried to iterate a product photo or a poster through ten rounds of "make the headline bigger, keep everything else," this is the feature list that matters.

Quality Upgrades

More natural lighting and richer textures
Sharper details overall
Better preservation of subjects from reference photos — people and products stay recognizable instead of drifting

Sketch: Draw Your Prompt

Type @Sketch in ChatGPT and a drawing canvas opens. Whatever you doodle becomes the visual reference for the final image — composition, layout, and placement come from your sketch, polish comes from the model. For anyone who thinks in rough layouts rather than prose prompts, this is the fastest prompt format yet.

Templates and Collaboration

Templates for popular formats — poster, merch, flyers, product photos — so the blank canvas problem disappears
Comments on images: drop a comment directly on a region to focus the next edit
Share with prompt: send an image with its generating prompt attached, so teammates can recreate the style with their own assets

API: Flare vs Sunburst

OpenAI split the API into two variants with a clear split:

Variant	Positioning	Latency	Best For
GPT-Image-2.5 Flare	Default choice for most apps	50% lower than GPT-Image-2	Social content, product exploration, visual search, rapid prototyping, high-volume generation
GPT-Image-2.5 Sunburst	Extra precision	Longer generation	Production ad creative, polished product imagery, workflows needing tight cross-edit control

Early external evaluation backs up Flare's speed claims: Manus measured it delivering high-quality images 2 to 4 times faster than GPT-Image-2. Both variants are available in the API today.

Pricing caveat: the launch announcement does not publish per-image rates — it links to OpenAI's pricing page. We won't guess numbers; check OpenAI's image generation pricing before budgeting. ChatGPT-side generation is included with existing plans, with usage limits that vary by tier.

Who Should Use It?

User Profile	Verdict
Marketers and social teams	Best fit — templates, fast iteration, and multi-edit consistency match campaign workflows
E-commerce sellers	Strong for product photos and merch mockups; reference-photo fidelity is the key upgrade
Developers (high volume)	Flare's 2-4x speed advantage at equal or better quality than GPT-Image-2
Designers needing pixel-perfect control	Try Sunburst, but expect longer waits; comment-on-image helps collaboration
Stylized/artistic generation	Test against Midjourney first — aesthetic-first generation is still its home turf

GPT Image 2.5 vs the Competition

Aspect	GPT Image 2.5	Midjourney	Adobe Firefly
Iterative editing	Excellent (multi-turn consistent)	Limited	Good
Conversational workflow	Native (ChatGPT)	No	No
Artistic style range	Good	Excellent	Good
Commercial safety (training data)	Vendor-specific	Vendor-specific	Adobe Stock-trained
Sketch-to-image input	Yes (@Sketch)	No	Limited
API variants for speed vs quality	Yes (Flare/Sunburst)	No	Limited

The pattern is clear: OpenAI is converging image generation with its assistant workflow — the model that lives where you already chat, edit, and collaborate. Dedicated art tools still lead on raw aesthetics, but the workflow integration gap is closing fast.

Summary

GPT Image 2.5 is a focused, practical upgrade rather than a flashy reinvention: 50% faster, sharper, and — most importantly — disciplined about preserving what you didn't ask to change across long editing sessions. Sketch and Templates attack the two hardest parts of real-world image work: communicating what you want, and starting at all.

The unknowns are pricing (unpublished at launch) and independent quality benchmarks (none yet). But for teams already in the ChatGPT ecosystem, it's the default image tool to evaluate first — and for API builders, the Flare/Sunburst split finally lets you buy speed and precision separately. For broader context on OpenAI's current lineup, see our GPT-6 Astra review.

What is GPT Image 2.5?

The headline change is speed: up to 50% lower generation latency than Images 2.0. But the more meaningful upgrades are in editing behavior — and that's where this model is aimed.