GPT Image 2.5 on Image 2 — OpenAI’s New Image Model

Available Now on Image 2

OpenAI released ChatGPT Images 2.5 (the GPT Image 2.5 generation) on September 8, 2026. On Image 2, GPT Image 2.5 is available in the model picker as Flare (fast everyday generation) and Sunburst (editing precision). You can still generate with GPT Image 2; prompts, references, and credits work the same way across models.

What Is GPT Image 2.5?

GPT Image 2.5 is OpenAI’s new state-of-the-art image generation and editing model line—productized in ChatGPT as ChatGPT Images 2.5, and exposed to developers as GPT‑Image‑2.5 Flare and GPT‑Image‑2.5 Sunburst in the Images API. It succeeds GPT Image 2 / ChatGPT Images 2.0 (April 2026) with sharper details, more natural lighting and richer textures, stronger preservation of people and objects in reference photos, more reliable multi-turn editing, and up to 50% lower latency versus Images 2.0.

OpenAI positions Images 2.5 for creative workflows where fidelity and control matter: transforming familiar subjects across new settings, updating one element without wrecking composition or brand treatment, and keeping quality stable across longer edit chains. In ChatGPT, the release also adds product tools such as Sketch (draw a rough guide and turn it into a finished image), templates for formats like posters and merch, on-image comments for focused edits, and the option to share prompts alongside images.

Image 2 (gptimage2.com) is an independent AI image workspace built around this model family. GPT Image 2.5 is live in the generator; you can also create with GPT Image 2, browse the community gallery, reuse references, and manage credits in one place.

Image 2 is not affiliated with OpenAI. Model names describe generation capabilities; the product brand is Image 2.

What’s New in OpenAI’s GPT Image 2.5

According to OpenAI’s announcement, ChatGPT Images 2.5 / GPT Image 2.5 focuses on:

1. Image fidelity from references — Subjects from reference photos stay more recognizable across styles, settings, and compositions; lighting and textures feel more natural; distinctive features are more likely to carry through.
2. Precision editing — Better at changing only what you asked for while keeping everything else the same—even with complex subjects and backgrounds. Useful for swapping a product, background, or line of copy without rebuilding the whole asset.
3. Multi-turn consistency — Editing instructions hold more reliably across multiple turns; earlier changes are less likely to unwind; quality degrades less as conversations get longer.
4. Intelligence and style — Stronger handling of complex visual instructions, more accurate real-world content in images, more complex layouts (including transparent backgrounds), and better adherence to requested visual styles.
5. Speed — Image generation latency reduced by up to 50% compared with Images 2.0, so iteration is faster.
6. ChatGPT creative tools — Sketch, templates, on-image comments, and prompt sharing (ChatGPT product features; Image 2 focuses on prompt + reference + gallery workflows).

OpenAI also notes continued safety measures (prompt and image checks, C2PA metadata, and invisible watermarking). Review outputs before publishing—AI images can still be imperfect.

GPT Image 2 vs GPT Image 2.5

| Focus | GPT Image 2 (Images 2.0) | GPT Image 2.5 (Images 2.5) | | ---------- | -------------------------------------------- | ------------------------------------------------------------------------------------------- | | Status | Available since April 2026 | Officially released Sept 8, 2026 by OpenAI | | Quality | Strong instruction following & on-image text | Sharper details, richer textures, more natural lighting | | References | Capable reference workflows | Better subject preservation from reference photos | | Editing | Solid with careful prompts | More precise local edits; stronger multi-turn consistency | | Latency | Baseline for Images 2.0 | Up to ~50% lower latency (OpenAI) | | API models | gpt-image-2 | gpt-image-2.5-flare (fast default) and gpt-image-2.5-sunburst (higher-control creative) | | On Image 2 | Available now | Available now (Flare + Sunburst) |

Flare vs Sunburst (API): OpenAI describes GPT‑Image‑2.5 Flare as the default for most apps—higher quality than GPT‑Image‑2 at lower latency, suited to creator/social content, product experiences, prototyping, and high-volume generation. GPT‑Image‑2.5 Sunburst targets premium workflows that need tighter control across edits (campaign creative, polished product imagery), typically with longer generation times. Image 2 exposes both as GPT Image 2.5 Flare and GPT Image 2.5 Sunburst in the model picker.

How GPT Image 2.5 Fits Image 2

1. Sign in and claim signup credits if you are new.
2. Open the generator on the homepage or /explore.
3. Choose GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, or GPT Image 2 in the model picker.
4. Write a prompt—or load one from the gallery with Use prompt.
5. Optionally add up to 16 reference images (especially useful given 2.5’s stronger reference fidelity).
6. Pick aspect ratio and resolution, confirm credits (typically five per run), and generate.
7. Review history, favorite keepers, and publish standouts.

Failed jobs show readable reasons so you can revise instead of guessing. Gallery favorites and creation history stay tied to your account across models.

Prompt Habits That Work on GPT Image 2.5

OpenAI’s release emphasizes fidelity, precision edits, and style control. These habits produce better results on GPT Image 2 and GPT Image 2.5:

Describe a picture, not a keyword stack.
Weak: poster, beautiful, 4k, masterpiece.
Strong: A vintage travel poster of a harbor town at golden hour, warm flat color fields, title across the top.

Quote exact on-image text.
Ask for a sign that reads "HARBOR LIGHT" in a bold serif. Quotation marks give the model a string to render.

Name layout hierarchy.
Headline large at the top, supporting lines in descending sizes, venue and date small at the base—especially useful as layouts get more complex.

One change per edit turn.
Change the label to matte black; keep lighting, subject, and framing exactly. This matches how Images 2.5 is designed to edit locally without drifting the rest of the frame.

Reuse strong references.
Because GPT Image 2.5 is better at preserving subjects from reference photos, start from a clear identity shot when brand or character consistency matters.

Who Benefits from GPT Image 2.5

Social and growth teams need faster iteration with sharper visuals. Marketers need precise edits that keep campaign composition intact. Product teams need reference-led variations that stay anchored to a hero shot. Indie creators need multi-turn polish without quality collapse. Developers building image features can target Flare for volume/speed or Sunburst for premium control—while Image 2 users pick either route in the model picker with a ready-made UI, gallery, and credits.

Credits, Pricing, and Expectations

Image 2 uses transparent credits for generation. Signup bonuses help you try GPT Image 2 and GPT Image 2.5; packs are on the pricing page. OpenAI publishes separate API token pricing for Flare/Sunburst; Image 2 credit costs are what you see in our UI before you click generate.

Always review AI images before client delivery. Prefer short, quoted text at larger sizes when typography accuracy is critical.

FAQ

Did OpenAI release GPT Image 2.5?
Yes. On September 8, 2026, OpenAI announced ChatGPT Images 2.5, with API models GPT‑Image‑2.5 Flare and GPT‑Image‑2.5 Sunburst. It is the successor generation to GPT Image 2 / Images 2.0.

Is GPT Image 2.5 available on Image 2 today?
Yes. Select GPT Image 2.5 Flare or GPT Image 2.5 Sunburst in the generator. GPT Image 2 remains available as well.

What’s the difference between Flare and Sunburst?
Per OpenAI: Flare is the fast, high-quality default for most applications (including ~50% lower latency vs GPT Image 2). Sunburst is for premium creative/editing workflows that need tighter multi-edit control. Both are selectable on Image 2.

Will my GPT Image 2 prompts work on GPT Image 2.5?
Generally yes—the prompt surface (text + optional image inputs) is the same family. Expect higher fidelity and stabler edits on the same instructions.

Is Image 2 part of OpenAI?
No. Image 2 is an independent service. GPT Image 2 / GPT Image 2.5 name underlying model capabilities.

Start Creating on Image 2

OpenAI’s GPT Image 2.5 release raises the quality and editability bar for AI images. On Image 2 you can use GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, or GPT Image 2 today—with gallery prompts, references, history, and clear credits.

Explore the gallery · Open the generator · View pricing · Sign up free