GPT Image 2.5 Sunburst: AI Image Editor Online
Bottom line
GPT Image 2.5 Sunburst editing: precision edits on up to 16 reference images, priced the same as Flare. Open the Image Edit tool, select GPT Image 2.5 Sunburst, upload your image, and the exact credit cost appears before the edit runs.
OpenAI released GPT Image 2.5 on 2026-09-08 in two variants that KIE exposes as four separate models on VdoBloom — Flare and Sunburst, each with its own text-to-image and image-to-image id. Sunburst is the precision variant, trading some of Flare's speed for edit precision and fidelity. Both accept the exact same input schema and — worth knowing before picking one — cost exactly the same on VdoBloom. Sunburst is not a premium tier; it trades latency, not price.
On editing, Sunburst shares Flare's headline capacity — up to 16 reference images per edit, six more than GPT Image 2's 10 — but is tuned for precision rather than turnaround. That suits composites drawing on many sources where the result needs to hold up under close inspection: matching lighting and perspective across references, keeping fine detail intact, rendering on-image text correctly.
It shares the same 13 aspect ratios, the same 1K/2K/4K resolution ladder at 3/5/8 credits, and the same 20,000-character prompt budget as Flare's editor. Unlike GPT Image 2's editor, there's no 1:1-at-4K restriction — only four exotic ratios (27:16, 16:27, 9:8, 8:9) cap at 1K, clamped automatically rather than failing the job.
It carries VdoBloom's STRICT content label and a 4.8 user rating, the highest among VdoBloom's GPT Image models. Credit cost is shown before every edit.
Sunburst vs Flare for editing
Identical 16-image capacity, identical price ladder — the difference is tuning. Sunburst favors edit precision and fidelity; Flare favors speed. VdoBloom users rate Sunburst 4.8 against Flare's 4.7. Since switching between them costs nothing extra, running a demanding edit on both and keeping the better result is a reasonable default.
What it costs on VdoBloom
3/5/8 credits for 1K/2K/4K, unchanged whether you fill all 16 reference slots or just two. A precision edit at 4K for final delivery costs 8 credits.
Common editing problems and how to avoid them
Most reports of noisy or inconsistent GPT Image 2.5 output trace back to an ambiguous instruction or a resolution silently clamped to 1K because the chosen aspect ratio was one of the four 1K-only ratios. Writing multi-part edits as separated, explicit steps and confirming the applied resolution before judging the result resolves most of what looks like a model flaw. A fully failed job is refunded automatically.
What creators use GPT Image 2.5 Sunburst for
- Precision composites from up to 16 reference images where lighting, perspective, and fine detail must match across sources.
- Edits with dense on-image text or exact layout instructions that need to render correctly the first time.
- Final-delivery edits at 4K (8 credits) after drafting the instruction on Flare or at a lower resolution.
How to use GPT Image 2.5 Sunburst on VdoBloom
- 1Open the Image Edit tool and select GPT Image 2.5 Sunburst.
- 2Upload the image(s) to edit and describe the change you want.
- 3Review the credit cost and generate — results appear in your library.
Frequently asked questions
GPT Image 2.5 Sunburst vs Flare for editing — which should I pick?
Sunburst trades some speed for edit precision and fidelity; Flare is faster. Both take the same inputs (up to 16 images) and cost the same (3/5/8 credits by resolution), so there's no downside to using Sunburst on a job where the edit needs to be exact.
How many reference images can Sunburst take in one edit?
Up to 16 — the same ceiling as Flare, and six more than GPT Image 2's 10-image limit.
How much does GPT Image 2.5 Sunburst editing cost?
3 credits at 1K, 5 at 2K, 8 at 4K — identical to Flare and to the text-to-image side.
Is it free to try?
10 free signup credits cover three 1K edits or one 2K edit with credits to spare.
Why does Sunburst rate higher than Flare on VdoBloom?
4.8 versus Flare's 4.7 — a small but consistent gap reflecting Sunburst's fidelity-focused tuning, at no extra cost.
GPT Image 2.5 vs GPT Image 2 for editing — what changed?
Reference capacity nearly doubled (16 vs 10 images), the 1:1-cannot-be-4K restriction is gone, and the four resolution-limited ratios clamp to 1K automatically instead of rejecting the job. Pricing is unchanged.
How strict is content filtering on Sunburst edits?
STRICT — VdoBloom's tightest tier, matching OpenAI's provider-side filtering, the same as Flare and GPT Image 2.
Some GPT Image 2.5 results look inconsistent or noisy — is Sunburst more reliable than Flare for that?
Sunburst is tuned for precision and fidelity, so it's a reasonable first move if a Flare result looks soft or noisy — at the same price, there's no cost penalty to trying it. Ambiguous prompts or a silently clamped resolution (check your aspect ratio isn't one of the four 1K-only ones) are the more common causes worth ruling out first.