How does it differ from GPT Image 2.5 Sunburst?

Flare is built for speed and is what OpenAI calls the default for most applications; Sunburst is built for edit precision and takes longer per image. They are two separate OpenAI model ids, and on RouteMux both bill per image and both support 2K / 4K, text-to-image and editing. Use Sunburst for images that will be inspected closely or retouched repeatedly, Flare for the rest.

How does it differ from GPT Image 2?

OpenAI says Flare produces higher-quality images than GPT Image 2 at roughly half the latency, and it adds the xhigh and max quality settings. On RouteMux, requests to the gpt-image-2 name are now served by Flare too, but that name only accepts quality up to high and rejects 2K / 4K sizes.

How do I get 2K or 4K?

Pass size 2048x2048, 2560x1440, 1440x2560 or the literal "2K"; for 4K pass 3840x2160, 2160x3840 or "4K". Endpoints and response shape stay the same; n must be 1, mask is not supported, and one image took about 33 to 45 seconds in our tests.

What happened to the plain gpt-image-2.5 name?

OpenAI never shipped a bare gpt-image-2.5 model id; the lineup is Flare and Sunburst. Calls to gpt-image-2.5 still work and are served by Flare, but new integrations should name Flare or Sunburst explicitly.

GPT Image 2.5 Flare

Save 70%Official

openai/gpt-image-2.5-flare

Also callable as gpt-image-2.5

OpenAI GPT Image 2.5 Flare — the fast tier of the 2.5 family and OpenAI's recommended default, billed per image.

Get an API key View code sample

Live model playground

Try GPT Image 2.5 Flare here

Tune the request, see the current price ceiling, and inspect the real response without leaving this model page.

asyncimage

1Configure input

Prompt

132 / 20,000

Source images for editing

Maximum estimated cost—

2Live resultExample

Input modalities: TextImage
Output modalities: Image
Capabilities: Image generation · Image editing
Upstream release: 2026-09-08
RouteMux name: openai/gpt-image-2.5-flare

What is GPT Image 2.5 Flare

GPT Image 2.5 Flare is one of the two image models OpenAI released on September 8, 2026 (snapshot gpt-image-2.5-flare-2026-09-08). OpenAI positions Flare for fast, high-quality everyday image generation and calls it the default choice for most applications, with higher-quality images than GPT Image 2 at roughly half the latency. Pick it for everyday generation, batch jobs, and anything where someone is waiting on the image. The other model, GPT Image 2.5 Sunburst, takes longer per image in exchange for edit precision. They are separate model ids, so you choose by name.

Compared with GPT Image 2, the 2.5 models add the xhigh and max quality settings (the full set is auto, low, medium, high, xhigh, max). On RouteMux billing is per image, not per token: prompt length, scene complexity and quality do not change the price; only the resolution tier does. At the standard tier each call returns one image, the size parameter picks the aspect ratio rather than the resolution, and the returned dimensions are the ones that count; they can exceed the request (a 1536x1024 request came back as 2304x1536 in our test).

For higher resolution, set size to a 2K or 4K value. 2K: 2048x2048, 2560x1440, 1440x2560, or simply 2K. 4K: 3840x2160, 2160x3840, or simply 4K. Measured at 16:9, 2K returns 2736×1536 and 4K returns 3840×2160. Text-to-image and image-to-image both work at 2K / 4K, with up to 10 reference images for edits; these tiers need n set to 1, do not support mask, and took about 33 to 45 seconds per image in our tests, possibly longer at peak times.

In every tier it speaks the OpenAI Images protocol: /v1/images/generations and /v1/images/edits take the same request shape as OpenAI, the call is synchronous, and the image comes back as base64 in data[0].b64_json.

Before you choose

How does it differ from GPT Image 2.5 Sunburst?: Flare is built for speed and is what OpenAI calls the default for most applications; Sunburst is built for edit precision and takes longer per image. They are two separate OpenAI model ids, and on RouteMux both bill per image and both support 2K / 4K, text-to-image and editing. Use Sunburst for images that will be inspected closely or retouched repeatedly, Flare for the rest.
How does it differ from GPT Image 2?: OpenAI says Flare produces higher-quality images than GPT Image 2 at roughly half the latency, and it adds the xhigh and max quality settings. On RouteMux, requests to the gpt-image-2 name are now served by Flare too, but that name only accepts quality up to high and rejects 2K / 4K sizes.
How do I get 2K or 4K?: Pass size 2048x2048, 2560x1440, 1440x2560 or the literal "2K"; for 4K pass 3840x2160, 2160x3840 or "4K". Endpoints and response shape stay the same; n must be 1, mask is not supported, and one image took about 33 to 45 seconds in our tests.
What happened to the plain gpt-image-2.5 name?: OpenAI never shipped a bare gpt-image-2.5 model id; the lineup is Flare and Sunburst. Calls to gpt-image-2.5 still work and are served by Flare, but new integrations should name Flare or Sunburst explicitly.

GPT Image 2.5 Flare pricing: Standard, 2K and 4K per image

Billed per image, not per token: Standard $0.0186 per image, 2K $0.08 per image, and 4K $0.12 per image. 2K and 4K support both text-to-image and image editing — just pass the matching size.

Prices verified 2026-10-23 · Charged only when an image is returned

StandardSave 70%

$0.0186/ image

Official ~~$0.062~~

✓ Text-to-image✓ Image editing

2KNew

$0.08/ image

Measured 2048×2048 · 2736×1536 · 1536×2736

✓ Text-to-image✓ Image editing

4KNew

$0.12/ image

Measured 3840×2160 · 2160×3840

✓ Text-to-image✓ Image editing

Estimate by image

Resolution

Images

Estimated cost

$8.00

Failed requests are free

How to get 2K / 4K

# set size to any 2K / 4K value from the table
curl https://api.routemux.com/v1/images/generations \
  -H "Authorization: Bearer $ROUTEMUX_API_KEY" \
  -d '{"model":"openai/gpt-image-2.5-flare",
       "prompt":"...", "size":"3840x2160"}'
# returns data[0].b64_json (PNG), same shape as Standard

Billing details

Price depends only on resolution — not on prompt length or quality
2K / 4K return one image per request (n must be 1); mask is not supported
A size not listed in the table returns 400 and is not charged
Standard-tier image editing also charges a small amount for input-image tokens
You pay only when an image is returned — failures, timeouts and blocked requests are free

GPT Image 2.5 Flare price per image (USD, verified 2026-10-23)
Tier	Measured output size	size value	Price per image	Vs. list price
Standard	—	—	$0.0186/ image	Official ~~$0.062~~Save 70%
2K	2048×2048 / 2736×1536 / 1536×2736	2048x2048, 2560x1440, 1440x2560	$0.08/ image	List price is per token; no published per-image price
4K	3840×2160 / 2160×3840	3840x2160, 2160x3840	$0.12/ image	List price is per token; no published per-image price

Key capabilities

The fast 2.5 tier: OpenAI calls it the default for most applications, with better quality than GPT Image 2 at about half the latency.
Flat price per image: One price per image in each resolution tier, independent of prompt length, scene complexity or quality.
2K / 4K output: Pick the tier with size. Text-to-image and edits both reach 2K and 4K; 4K at 16:9 measured 3840×2160.
Quality up to max: Adds xhigh and max over GPT Image 2, without changing the per-image price on RouteMux.

Limited time10% extra Credits on your first top-up — up to 200 bonus Credits

Applies to your first top-up within 72 hours of signing up. Minimum $20, once per account.

Interface and protocols

OpenAI Images

Use any of these IDs to call this model via the API.

RouteMux nameopenai/gpt-image-2.5-flare

Common namegpt-image-2.5

Replace the ROUTEMUX_KEY placeholder with your API key. Create one →

from openai import OpenAI

client = OpenAI(
    base_url="https://api.routemux.com/v1",
    api_key="<ROUTEMUX_KEY>",
)

response = client.images.generate(
    model="openai/gpt-image-2.5-flare",
    prompt="A red panda coding under the moonlight",
    n=1,
)
print(response.data[0].url)

base_url: https://api.routemux.com/v1

Performance (last 30 days)

693 requests · last 30 days

Total usage1.10M

P50 first-token latency32.68s

Avg throughputInsufficient data (693 requests)

Success rate100.0%

Uptime · 24htested via openai_images

100.00%

Success rate · 60d · from real requests

GPT Image 2.5 Flare FAQ

›What is GPT Image 2.5 Flare?

OpenAI GPT Image 2.5 Flare — the fast tier of the 2.5 family and OpenAI's recommended default, billed per image.

›How much does the GPT Image 2.5 Flare API cost?

Through RouteMux, GPT Image 2.5 Flare is priced at $0.0186, below the official list price.

›How do I call GPT Image 2.5 Flare via API?

GPT Image 2.5 Flare is OpenAI-compatible: point your base URL at RouteMux and set the model field to openai/gpt-image-2.5-flare — no code changes needed.

›Which input and output modalities does GPT Image 2.5 Flare support?

GPT Image 2.5 Flare accepts text, image as input and produces image as output.

›When was GPT Image 2.5 Flare released?

GPT Image 2.5 Flare was released on September 8, 2026.

Sources

Reviewed by the RouteMux team on 2026-10-24

View as Markdown

What is GPT Image 2.5 Flare

GPT Image 2.5 Flare pricing: Standard, 2K and 4K per image

Estimate by image

How to get 2K / 4K

Billing details

Key capabilities

Interface and protocols

Performance (last 30 days)

Related models

GPT Image 2.5 Flare FAQ

Sources