ChatGPT Images 2.5: Presentation Visuals, Access, and Alternatives (2026)
ChatGPT Images 2.5 is OpenAI's image model announced on September 8, 2026. It generates images faster than the previous version, edits more precisely, and holds detail steady across several rounds of changes. This guide covers what changed, who can use it, what it means for slide visuals, and which alternatives are worth knowing.
What changed in this release
OpenAI's announcement page opens with a number worth pausing on. Every week, people create "more than 3 billion images across ChatGPT Images and the GPT‑Image models in the API."
Four changes stand out for anyone who makes visuals for work.
Faster generation. OpenAI writes that it has "reduced image generation latency by up to 50% compared with Images 2.0." For iterative work, that difference compounds.
More reliable editing. The model "is better at editing only what you've asked for, while keeping the rest of the details the same." Backgrounds and composition survive a targeted change.
OpenAI describes the same behaviour for developers in concrete terms. The API can update a single element, "such as a product, background, or piece of copy." The subject, composition, and brand treatment around it stay put. For a deck, the equivalent is swapping one figure in a diagram without redrawing the diagram.
Consistency across turns. Earlier edits "are more likely to stay consistent, and each new edit builds on the work you've already done without degrading image quality over time."
Complex layouts. The page notes the model "can handle more complex layouts including transparent backgrounds."
Three product features ship alongside the model. Sketch lets you draw directly in ChatGPT as a visual guide, reachable by typing @Sketch. Templates give you a starting point for common formats, described on the page as choices like "Poster" or "Merch." You can also place comments directly on an image, and share the prompt behind a result so someone else can run with it.
What this means for presentation visuals
The announcement is unusually explicit about deck work. OpenAI spells out where the intelligence and style gains land. They matter "when generating a series of on-brand creative assets, UI concepts that preserve a given hierarchy, or presentation visuals that fit a defined structure."
That phrase, a defined structure, is the part that changes day-to-day work. A deck is not one image. It is eight or twelve that have to look like siblings, and the older failure mode was drift: slide nine no longer matched slide two.
The multi-turn claim addresses the same failure from another angle. OpenAI writes that earlier changes "are more likely to stay consistent." Each new edit builds on previous work "without degrading image quality over time." Anyone who has regenerated a deck illustration five times knows why that sentence matters.
Transparent-background support pushes in the same direction. Lucky Liao of Manus is quoted on the page to that effect. The improvement makes the model "an excellent fit for creative work on Manus, from brand assets to presentations and websites."
So the honest summary is narrow but real. ChatGPT Images 2.5 makes a set of visuals easier to keep coherent. Which visuals your deck needs is a question the announcement does not take up.
Who can access it today
| Where | What you get | Notes |
|---|---|---|
| ChatGPT | Images 2.5 | Rolling out across all tiers |
| ChatGPT Work | Images 2.5 | Same rollout |
| Codex | Images 2.5 | Same rollout |
| Platforms | Desktop, mobile, and web | Stated on the announcement page |
| API | GPT‑Image‑2.5 Flare and Sunburst | Two separate models |
OpenAI states that Images 2.5 "is rolling out today to ChatGPT, ChatGPT Work, and Codex users across all tiers on desktop, mobile, and web." Availability by tier is therefore broad, which is a change from releases that started with paid plans only.
On provenance, the page says OpenAI continues "to use C2PA metadata and invisible watermarking to help identify images made with our tools." If your team publishes generated visuals in regulated material, that detail belongs in your review checklist.
Flare or Sunburst: which API model fits which job
Developers get two models rather than one, and the split is about speed against control.
| Model | Positioned for | Official description |
|---|---|---|
| GPT‑Image‑2.5 Flare | Most applications | "the default choice for most applications, delivering higher-quality images than GPT‑Image‑2 at 50% lower latency" |
| GPT‑Image‑2.5 Sunburst | Premium visual work | "built for premium visual workflows that benefit from tighter control across edits" |
The page suggests Flare for "creator and social content to product experiences, visual search, rapid image prototyping, and high-volume generation." Sunburst is aimed at "production-ready campaign creative or polished product imagery," and takes longer to generate.
Adobe's Matt Chotin is quoted on the same page. The comment describes the pair as "bringing faster generation and resolution consistency that keep images sharp and photorealistic through every refinement."
Where a faster image model stops helping
The announcement describes Images 2.5 entirely in terms of what appears in the frame: lighting, textures, subject fidelity, layout, editing precision. That is the job it is built for, and the page makes no claim beyond it.
A data slide has a second job. The bar has to be the right height, and the label has to match the row it came from. Someone in the room may also ask where the number came from. Those answers live in a spreadsheet, not in a rendering.
This is where the two kinds of tool divide. Powerdrill Bloom starts from the file. You upload the export, ask in natural language for the cut you want, and the platform returns the figure with its source attached. The product page puts it plainly: every number "comes back with the page, the row and the figure behind it."
The pricing page describes the same output shape, offering a grounded answer with "charts, tables, and exports." Free covers everyday exploration with 1,000 daily refreshed credits, plus basic slides, docs, sheets, and images. Pro at $13.27 per month adds deeper analysis, richer outputs, and 20 scheduled tasks. Details for each tier are listed on the pricing page.
Neither tool replaces the other. One makes the picture; the other makes the picture true.
Alternatives worth knowing
Nano Banana. Strong on the same job most teams actually want, which is a visual that drops into a deck. Our guide to Nano Banana covers where it fits.
Adobe Firefly. Adobe's comment on the announcement page confirms the new GPT‑Image‑2.5 models are now available inside Firefly. That makes the choice less about the model and more about the editor around it.
Runway. Jamie Umpherson, its chief creative officer, is quoted saying the model "fits seamlessly into how creators already work in Runway."
Chart and infographic generators. If the visual you need is a chart rather than an illustration, a purpose-built tool is a better fit. See our roundup of AI infographic generators, or the AI infographic generator page for the data-driven route.
Conclusion
ChatGPT Images 2.5 is a solid step for anyone producing a run of visuals that has to hold together. Faster generation, tighter editing, and better multi-turn consistency all point at the same use case: a set, not a single picture.
Just keep the division of labour straight. Use an image model for the look, and something that reads your actual file for the numbers on the slide. Our walkthrough on adding generated images, graphs, and infographics to slides covers the handoff between the two.
Have a data file that needs to become a deck? Try Powerdrill Bloom and start from the numbers.
Frequently asked questions
What is ChatGPT Images 2.5?
It is OpenAI's image generation model announced on September 8, 2026. OpenAI describes it as bringing sharper details, more precise editing, and faster generation, with latency reduced by up to 50% compared with Images 2.0.
Who can use Images 2.5?
OpenAI states it is rolling out to ChatGPT, ChatGPT Work, and Codex users across all tiers, on desktop, mobile, and web. Developers get two separate models through the API.
What is the difference between Flare and Sunburst?
Flare is described as the default choice for most applications, with 50% lower latency than GPT‑Image‑2. Sunburst is built for premium visual workflows needing tighter control, and takes longer to generate.
Can it make charts from my spreadsheet?
The announcement describes image generation, not data analysis. For a chart tied to your own rows, a tool that reads the file is the better route, such as an AI chart generator.
Are generated images marked in any way?
Yes. OpenAI says it continues to use C2PA metadata and invisible watermarking to help identify images made with its tools.