gpt-image-2.5-flaregpt-image-2.5-flare is the tier you build a pipeline on. It carries the whole GPT Image 2.5 feature set — point at the part of a picture you don't like and only that part changes, the same character or product survives a long chain of edits instead of drifting after a handful, and transparent PNG is finally native rather than a flat backdrop you key out yourself — but it returns in roughly half the time GPT Image 2 took, and about twice as fast as Sunburst on the same prompt. Quality defaults to medium, which is the right setting for social posts, catalogue tiles, rapid prototyping and anything you generate in volume. Sizes are native at 1K, 2K and 4K, in 1:1, 3:4, 4:3, 9:16, 16:9 and 21:9. Worth knowing before you pick a tier: Flare is priced by resolution × quality across fifteen cells, from $0.008 at 1K low to $0.3 at 4K max. At 1K and 2K Sunburst bills exactly the same cells, so at those sizes you are choosing speed, not budget. At 4K the two part ways — Sunburst runs $0.02 / $0.085 / $0.16 / $0.28 / $0.47 from low to max against $0.02 / $0.045 / $0.08 / $0.17 / $0.3 here — which makes Flare the natural home for 4K volume work. The defaults differ as well (medium here, high on Sunburst), and quality is a price axis. Sketch input, the creation templates and comment-on-image editing are ChatGPT features at launch and are not part of the API.
Generated image will appear here
Enter a prompt and click Generate
Priced on two axes, resolution and quality — fifteen cells. 1K: $0.008 / $0.020 / $0.03 / $0.05 / $0.08. 2K: $0.020 / $0.025 / $0.04 / $0.08 / $0.15. 4K: $0.02 / $0.045 / $0.08 / $0.17 / $0.3 (low to max). You pay for the detail you ask for. The tier is decided by the total pixel count of the image you actually get back — 1.7MP or less bills as 1K, 1.7-4.6MP as 2K, above 4.6MP as 4K — not by its longest edge and not by the parameter you sent. A 2K image at 2720x1536 is 4.18MP, so it bills as 2K even though its long edge exceeds 2048. Transparent background and reference images are included; failed requests are not charged.
Mark the part you want changed and the rest of the frame is left alone — swap an outfit without the face shifting, rewrite on-image text without the background moving. This is the change most people notice first.
The same character or product holds together over a long chain of edits instead of drifting after a few. Keeping a set on-model no longer depends on structural prompts, reference images and fixed-lighting clauses.
Real alpha, not a flat colour you have to key out. GPT Image 2 could not do this, so a transparent-background request is the simplest way to tell the two models apart.
At 1K and 2K, Flare and Sunburst bill the same cell for the same resolution and quality, so the choice is speed against precision. At 4K Sunburst is priced higher: $0.02 / $0.085 / $0.16 / $0.28 / $0.47 from low to max, against $0.02 / $0.045 / $0.08 / $0.17 / $0.3 here. The defaults differ too (medium here, high on Sunburst). Pick Flare when turnaround and volume decide the job, Sunburst when someone is going to zoom in.
About half the latency of GPT Image 2 and roughly twice the speed of Sunburst on the same prompt, which is what makes it viable for catalogue runs, social pipelines and prototyping loops where you generate hundreds of variants.
Sketch (draw a rough layout and let the model read it), the 15 creation templates and commenting directly on an image are ChatGPT features at launch. They are not exposed through the API — worth knowing before you build against them.
GPT Image 2.5 Flare is a Image Generation API provided by OpenAI. gpt-image-2.5-flare is the tier you build a pipeline on. It carries the whole GPT Image 2.5 feature set — point at the part of a picture you don't like and only that part changes, the same character or product survives a long chain of edits instead of drifting after a handful, and transparent PNG is finally native rather than a flat backdrop you key out yourself — but it returns in roughly half the time GPT Image 2 took, and about twice as fast as Sunburst on the same prompt. Quality defaults to medium, which is the right setting for social posts, catalogue tiles, rapid prototyping and anything you generate in volume. Sizes are native at 1K, 2K and 4K, in 1:1, 3:4, 4:3, 9:16, 16:9 and 21:9. Worth knowing before you pick a tier: Flare is priced by resolution × quality across fifteen cells, from $0.008 at 1K low to $0.3 at 4K max. At 1K and 2K Sunburst bills exactly the same cells, so at those sizes you are choosing speed, not budget. At 4K the two part ways — Sunburst runs $0.02 / $0.085 / $0.16 / $0.28 / $0.47 from low to max against $0.02 / $0.045 / $0.08 / $0.17 / $0.3 here — which makes Flare the natural home for 4K volume work. The defaults differ as well (medium here, high on Sunburst), and quality is a price axis. Sketch input, the creation templates and comment-on-image editing are ChatGPT features at launch and are not part of the API. Through APIMODELS platform, you can access this model via a unified API with transparent pay-as-you-go pricing. Current pricing: 1K-low: $0.008, 1K-medium: $0.02, 1K-high: $0.03, 1K-xhigh: $0.05, 1K-max: $0.08, 2K-low: $0.02, 2K-medium: $0.025, 2K-high: $0.04, 2K-xhigh: $0.08, 2K-max: $0.15, 4K-low: $0.02, 4K-medium: $0.045, 4K-high: $0.08, 4K-xhigh: $0.17, 4K-max: $0.3.
One hero shot, then colourways, materials and packaging text edited in place — the product stays the same object across the whole set instead of being re-imagined each time.
background=transparent returns a real alpha PNG at 1K, 2K or 4K, so catalogue tiles, stickers and compositing layers skip the keying step entirely.
A mascot, model or spokesperson that survives dozens of edits — different outfits, props and scenes without the face drifting. Reference images are free, so anchor the set once and iterate.
Change the headline on a poster or the label on a bottle and the composition, lighting and typography around it stay put — the edit is targeted at the words, not the frame.
About half the wait of GPT Image 2, and roughly half of Sunburst on the same prompt — which turns a few hundred catalogue tiles or social variants into an afternoon rather than a week. Same price per image as Sunburst at the same quality, so the speed costs nothing.
GPT Image 2.5 Flare is available through APIMODELS at: 1K-low: $0.008, 1K-medium: $0.02, 1K-high: $0.03, 1K-xhigh: $0.05, 1K-max: $0.08, 2K-low: $0.02, 2K-medium: $0.025, 2K-high: $0.04, 2K-xhigh: $0.08, 2K-max: $0.15, 4K-low: $0.02, 4K-medium: $0.045, 4K-high: $0.08, 4K-xhigh: $0.17, 4K-max: $0.3. Billing is pay-as-you-go — you only pay for what you generate.
Sign up at APIMODELS, get your API key, and call our unified API endpoint. We provide detailed API documentation with code examples in cURL, Python, and Node.js.
APIMODELS offers the same GPT Image 2.5 Flare model through our aggregation platform. We provide a unified API interface so you do not need separate accounts for each provider - one API key to access all models.
Flare is the default tier: full 2.5 quality at roughly half the latency of GPT Image 2, rendering at medium unless you pass quality — the right pick for social content, e-commerce and anything high-volume. Sunburst is the highest-fidelity tier, renders at high by default, takes longer per image, and is meant for premium commercial visuals. Both accept identical parameters and share one price grid at 1K and 2K; at 4K Sunburst is priced higher (see the next answer). Beyond that, the difference is the default quality and the upstream fidelity tuning. Start with flare; when you need more, switch the model to sunburst or simply pass quality:"high" on the same request.
Priced on two axes — resolution AND quality — fifteen cells in all. 1K: $0.008 / $0.020 / $0.03 / $0.05 / $0.08. 2K: $0.020 / $0.025 / $0.04 / $0.08 / $0.15. 4K: $0.02 / $0.045 / $0.08 / $0.17 / $0.3 (low / medium / high / xhigh / max). You pay for the detail you actually ask for: a layout check at 1K low is $0.008, and you only reach for max when the image is going to print. The tier is decided by the total pixel count of the image you actually get back — 1.7MP or less bills as 1K, 1.7-4.6MP as 2K, above 4.6MP as 4K — not by its longest edge and not by the parameter you sent. A 2K image at 2720x1536 is 4.18MP, so it bills as 2K even though its long edge exceeds 2048 (with size auto or omitted, the output decides). Reference images and transparent background cost nothing extra, and failed requests are not charged. With quality omitted or auto you get the model's default — medium for flare, high for sunburst. At 1K and 2K both tiers cost the same at a given quality, so there you are choosing speed, not price; at 4K sunburst has its own, higher row: $0.02 / $0.085 / $0.16 / $0.28 / $0.47 from low to max.
It depends on the tier you use. OpenAI bills 2.5 per output token ($30 per million image tokens), so higher quality costs more — the same 4K image is about $0.011 at low and about $0.40 at max, a 36x spread. We bill on the same two axes, so every cell compares directly: **we are more affordable everywhere from high upward** — 1K high $0.03 against about $0.053, 2K high $0.04 against about $0.107, 4K high $0.08 against about $0.100, 4K xhigh $0.17 against about $0.178 and 4K max $0.3 against about $0.40; 2K medium is a little lower too ($0.025 against about $0.027). The OpenAI list rate is lower only at the low tiers and at 1K / 4K medium (2K low about $0.012 against our $0.020; 1K medium about $0.013 against our $0.020). The real difference is not the unit price: no OpenAI organisation verification, direct access from mainland China, one key for 110+ models on the site, and failed generations are not charged.
Transparent backgrounds do: background=transparent with output_format=png returns a real alpha PNG, verified at 1K, 2K and 4K (52-67% transparent pixels on product shots). Masks are gone. A mask is only worth anything as a hard boundary, and none of the upstream channels we tested treats it as one — they accept it, return 200 and repaint the protected region anyway. Rather than take a parameter we would silently ignore, a request carrying mask now returns 400. Keep an edit local through the prompt instead: name the object to change and say the rest stays, as in 'change only the red circle to blue, leave everything else unchanged' — in our tests only the circle changed.
All three tiers are native renders, not upscales — we checked with a detail-energy test (downscale by half, upscale back, measure what is lost), and the 2K and 4K outputs carry as much full-resolution high-frequency detail as 1K. On size: all three tiers are delivered exactly as requested, with no snapping — 1024×1024 at 1K, 2720×1536 at 2K (1536×2720 for 9:16) and 3840×2160 at 4K (2160×3840 for 9:16); 16 generations across both orientations on 2026-09-22 all came back at the requested size.
Production medians over the last seven days (external traffic, successful calls, 2026-09-19): about 45s at 1K (p90 around 65s), about 25s for flare and 40s for sunburst at 2K, about 35s for sunburst and 50s for flare at 4K; reference-image edits 15-50s. 1K is currently slower than 2K — a channel difference, not a model one. For synchronous calls set client timeouts to 180s or more, or use the async taskId polling / webhook contract instead.
On APIMODELS, GPT Image 2.5 Flare runs alongside 60+ models on one API key and one balance, so choosing is about fit, not lock-in. It supports Text to Image, Precise region edit — mark a spot, only that changes, Multi-turn consistency without prompt scaffolding, Native transparent PNG, Sharper detail, faster generation, 1:1 / 3:4 / 4:3 / 9:16 / 16:9 / 21:9, Flare (default, ~2x faster) and Sunburst (highest fidelity), and you can weigh it on price and capability against other Image Generation models, then switch by changing a single model-name string — no new account or integration. Browse every Image Generation option with live pricing at apimodels.app/models.
GPT Image 2.5 Flare supports: Text to Image, Precise region edit — mark a spot, only that changes, Multi-turn consistency without prompt scaffolding, Native transparent PNG, Sharper detail, faster generation, 1:1 / 3:4 / 4:3 / 9:16 / 16:9 / 21:9, Flare (default, ~2x faster) and Sunburst (highest fidelity). See the APIMODELS docs for full parameters and call examples.
Yes. APIMODELS exposes GPT Image 2.5 Flare through a single unified API and one key — no separate provider accounts, and no need to handle each provider's regional network access yourself.
We support Stripe (Visa, Mastercard, and other international cards) and Alipay. Credits are available instantly after payment.
A lot of people land here looking for a way to fix one line of wording inside a picture, not for an API key. That job has a tool of its own: EditTextImage, which we built on top of our image models and which needs nothing but the image, the current wording and the replacement. Start with edit text in a JPG if your file is a photo or an export, or read edit text in an image with the same font for how the original typography is matched.
Prompts shared by their authors — copy and adapt them. Each one credits its author and links back to the original post.
Weathered sailor on a fishing boat
Create a photorealistic candid photograph of an elderly sailor standing on a small fishing boat. He has weathered skin with visible wrinkles, pores, and sun texture, and a few faded traditional sailor tattoos on his arms. He is calmly adjusting a net while his dog sits nearby on the deck. Shot like a 35mm film photograph, medium close-up at eye level, using a 50mm lens. Soft coastal daylight, shallow depth of field, subtle film grain, natural color balance. The image should feel honest and unposed, with real skin texture, worn materials, and everyday detail. No glamorization, no heavy retouching.
by OpenAI
Automatic coffee machine workflow infographic
Create a detailed Infographic of the functioning and flow of an automatic coffee machine like a Jura. From bean basket, to grinding, to scale, water tank, boiler, etc. I'd like to understand technically and visually the flow.
by OpenAI
Thread streetwear ad with exact typography
Give me a cool in culture ad / fashion shot for a brand called Thread. It's a hip young street brand. The ad shows a group of friends hanging out together with the tagline "Yours to Create." Make it feel like a polished campaign image for a youth streetwear audience: stylish, contemporary, energetic, and tasteful. Use clean composition, strong color direction, natural poses, and premium fashion photography cues. Render the tagline exactly once, clearly and legibly, integrated into the ad layout. No extra text, no watermarks, no unrelated logos.
by OpenAI
We curate copy-ready prompt libraries — every entry shows its full text and a sample result, ready to adapt.
How to get access, regional availability, and how this model compares with its alternatives.