API Key Setup12 min readPublished: 2026-09-09Updated: 2026-09-09

ChatGPT Images 2.5 API Guide: Flare vs Sunburst, Pricing & Code

Understand ChatGPT Images 2.5 and its two API models: GPT-Image-2.5 Flare for fast everyday generation and Sunburst for precision editing. Compare pricing, model IDs, features, and working API examples.

By Heizi· Founder & Editor· Published: 2026-09-09Official sources checked · 2026-09-09

Verification: Model IDs, pricing, availability, features, and code patterns were checked against the OpenAI sources listed below. Get Model Key has not independently run Flare or Sunburst through the API, so output quality and latency claims are attributed to OpenAI and should be verified in your own project.

Create an OpenAI API key

ChatGPT Images access and API usage are separate. API image generation is billed through your OpenAI Platform project.

OpenAI ChatGPT Images 2.5 cover comparing Flare generation with precise Sunburst image editing

What OpenAI released with Images 2.5

OpenAI announced ChatGPT Images 2.5 on September 8, 2026. The product update brings sharper details, more natural lighting and textures, more reliable targeted edits, and better consistency across multiple editing turns. OpenAI says generation latency is reduced by up to 50% compared with Images 2.0.

The naming matters: ChatGPT Images 2.5 is the experience inside ChatGPT, while developers choose between two API model IDs—gpt-image-2.5-flare and gpt-image-2.5-sunburst. There is no single documented gpt-image-2.5 model ID in the launch materials.

Product name: ChatGPT Images 2.5
Fast API model: gpt-image-2.5-flare
Precision API model: gpt-image-2.5-sunburst
Release date: September 8, 2026

GPT-Image-2.5 Flare vs Sunburst

Flare is the default choice for most applications. OpenAI positions it as the fastest high-quality model for everyday generation, creator tools, social content, product experiences, visual search, rapid prototyping, and high-volume workloads.

Sunburst is the premium choice when editing precision matters more than turnaround time. It is designed for production campaign assets, polished product imagery, and workflows where requested changes must preserve the rest of the composition.

Model	Priority	Best for	Snapshot
gpt-image-2.5-flare	Speed + quality	Everyday and high-volume generation	gpt-image-2.5-flare-2026-09-08
gpt-image-2.5-sunburst	Editing precision	Premium creative and product editing	gpt-image-2.5-sunburst-2026-09-08

GPT-Image-2.5 API pricing

Both official model pages list the same token rates: $5.00 per million text-input tokens, $1.25 cached text input, $8.00 image input, $2.00 cached image input, and $30.00 image output. Text output is not billed because these models output images.

Token rates are not a fixed price per picture. Final cost depends on the request, source images, dimensions, quality, and output token usage. OpenAI also notes that the GPT-Image-2 calculator does not estimate GPT-Image-2.5 token consumption, so measure real requests in your usage dashboard before forecasting production spend.

Token rates shared by Flare and Sunburst, checked September 9, 2026
Token type	Price per 1M tokens
Text input	$5.00
Cached text input	$1.25
Image input	$8.00
Cached image input	$2.00
Image output	$30.00

Generate your first image with Flare

Create an OpenAI API key, keep it in the OPENAI_API_KEY environment variable, and start with Flare. The model supports low, medium, high, xhigh, max, and auto quality settings. Use auto while evaluating unless your workflow has a clear latency or fidelity target.

The example writes the returned base64 image to a local PNG file. Keep API calls on your server; exposing the key in frontend JavaScript lets anyone spend against your account.

python

from openai import OpenAI
import base64

client = OpenAI()
result = client.images.generate(
    model="gpt-image-2.5-flare",
    prompt="A clean editorial product photo of a translucent blue desk lamp",
    size="1536x1024",
    quality="auto",
)

image_bytes = base64.b64decode(result.data[0].b64_json)
with open("flare-output.png", "wb") as image_file:
    image_file.write(image_bytes)

Python image-generation example using the Flare model ID.

Use Sunburst for precision editing

Sunburst accepts text and image input and is designed to preserve subjects, composition, and visual identity across targeted edits. State both the requested change and the invariants that must remain untouched. A vague prompt such as 'make it better' gives the model no reliable edit boundary.

For repeatable production output, pin the dated snapshot and keep your prompt, source asset, quality, and output settings in an evaluation fixture. Compare both visual quality and the number of manual corrections required.

python

from openai import OpenAI
import base64

client = OpenAI()
with open("product.png", "rb") as source_image:
    result = client.images.edit(
        model="gpt-image-2.5-sunburst-2026-09-08",
        image=source_image,
        prompt=(
            "Replace only the background with a warm studio gradient. "
            "Keep the product, label, proportions, camera angle, and shadows unchanged."
        ),
        quality="high",
    )

with open("sunburst-edit.png", "wb") as output_file:
    output_file.write(base64.b64decode(result.data[0].b64_json))

Precision-edit example using the dated Sunburst snapshot.

What is new inside ChatGPT Images 2.5?

Images 2.5 is rolling out across ChatGPT, ChatGPT Work, and Codex on desktop, mobile, and web. Inside ChatGPT, Sketch lets you draw a visual reference, templates provide starting formats such as flyers and product photos, comments target edits to a specific part of an image, and shared prompts let other people recreate an idea with their own details.

Those ChatGPT interface features are not the same as API parameters. The API exposes generation and editing models; your application must build its own canvas, comments, version history, storage, and collaboration layer.

Prompting and evaluation checklist

Structure prompts around subject, environment, composition, lighting, materials, exact text, and constraints. For edits, explicitly repeat what must not change. When text accuracy matters, put the exact copy in quotation marks and validate the rendered result instead of assuming it is correct.

Evaluate on your own asset set. Track prompt success rate, edit preservation, text accuracy, identity consistency where permitted, latency, token cost, and human correction time. Flare and Sunburst have the same listed token rates, so model selection should focus on workflow results rather than a presumed price difference.

Use Flare as the default baseline
Use Sunburst when edit precision reduces costly rework
Pin dated snapshots for repeatable production evaluations
Keep API keys server-side and set project spending limits
Preserve C2PA metadata where your delivery pipeline allows it

Safety, provenance, and production controls

OpenAI says Images 2.5 continues to use prompt and image safety checks, C2PA metadata, and invisible watermarking. Do not remove provenance metadata merely to make generated media appear human-made, and do not assume these controls replace your own moderation or rights review.

For production, record the model ID, snapshot, prompt version, source-asset rights, user consent, and moderation outcome. Require human approval for public campaigns, sensitive edits, identity-related work, and any output that could be mistaken for documentary evidence.

FAQ

What are the GPT-Image-2.5 API model IDs?

Use gpt-image-2.5-flare for fast everyday generation or gpt-image-2.5-sunburst for precision editing. Dated snapshots ending in 2026-09-08 are also available.

Is ChatGPT Images 2.5 the same as GPT-Image-2.5?

They belong to the same release, but the names refer to different surfaces. ChatGPT Images 2.5 is the ChatGPT product experience; the API exposes Flare and Sunburst model IDs.

Should I use Flare or Sunburst?

Start with Flare for most generation workloads. Test Sunburst when preserving exact details across edits is more important than speed.

How much does GPT-Image-2.5 cost per image?

There is no single fixed per-image price in the official model pages. Cost depends on text input, image input, output tokens, dimensions, quality, and the request. Measure representative calls in your API usage dashboard.

Is Images 2.5 available in ChatGPT?

OpenAI says it is rolling out to ChatGPT, ChatGPT Work, and Codex users across all tiers on desktop, mobile, and web. Feature visibility can still vary during rollout.