GPT Image 2.5:Flare, Sunburst, API pricing, and editing

What is GPT Image 2.5?

ChatGPT Images 2.5 is OpenAI’s September 2026 image generation update. OpenAI highlights sharper detail, more natural lighting and texture, faster generation, stronger subject preservation, and more reliable edits across multiple turns. This page is an independent LAYSO guide based on the linked official material.

The ChatGPT product name and API model names are related but not interchangeable. Developers choose either GPT-Image-2.5 Flare or GPT-Image-2.5 Sunburst. Both accept text and image inputs and produce images; their main tradeoff is speed versus editing precision.

Source: Introducing ChatGPT Images 2.5.

GPT Image 2.5 Flare vs Sunburst

OpenAI positions Flare as the default choice for most applications and Sunburst as the premium option for workflows that benefit from tighter control. Test both with the same prompt, input images, quality setting, and acceptance rubric before choosing a production default.

GPT-Image-2.5 Flare

gpt-image-2.5-flare

Fast, high-quality everyday generation and editing.

Product experiences, creator tools, rapid prototypes, visual search, and higher-volume generation.

Official model pageDownload icon

GPT-Image-2.5 Sunburst

gpt-image-2.5-sunburst

Maximum editing precision for premium visual workflows.

Campaign creative, polished product imagery, and iterative edits where precise control matters more than latency.

Official model pageDownload icon

Source: OpenAI image generation guide.

Reference fidelity and subject preservation

A useful image edit changes the requested element without quietly rebuilding everything else. OpenAI reports improved preservation of recognizable subjects, composition, lighting, texture, and other details from reference images. The paired official example below shows the type of constrained transformation teams should evaluate with their own material.

Source: Introducing ChatGPT Images 2.5.

Precision editing and multi-turn consistency

Images 2.5 is designed to follow focused edit instructions while preserving unmentioned details. OpenAI also reports better consistency over longer editing conversations, so a new change is less likely to erase earlier decisions or degrade the image.

Source: Introducing ChatGPT Images 2.5.

GPT Image 2.5 API: models, endpoints, and first request

Use the Image API for a direct generation or edit request. Use the Responses API when image generation belongs inside a conversation or multi-step workflow. Both APIs expose quality, size, format, and compression controls. OpenAI documents the quality values low, medium, high, xhigh, max, and auto for both 2.5 models.

The example below uses the official OpenAI endpoint and requires an OpenAI API key. It does not send a request from this page.

import OpenAI from "openai";
import fs from "node:fs";

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
});

const result = await openai.images.generate({
  model: "gpt-image-2.5-flare",
  prompt: "A premium product photograph of a ceramic lamp, warm studio light, neutral background",
  quality: "high",
});

const bytes = Buffer.from(result.data[0].b64_json, "base64");
fs.writeFileSync("lamp.png", bytes);

Source: OpenAI image generation guide.

GPT Image 2.5 API pricing

OpenAI lists the following USD rates per one million tokens for both Flare and Sunburst. These are OpenAI reference rates, not a LAYSO quote. The published GPT Image 2 calculator does not estimate GPT Image 2.5 token consumption, so measure real requests before projecting spend.

Text input

$5.00

Cached text input

$1.25

Image input

$8.00

Cached image input

$2.00

Image output

$30.00

Source: GPT-Image-2.5 Flare model documentation.

A practical production checklist

The following is LAYSO editorial guidance, not an additional OpenAI model specification.

1. Choose the model by evidence. Compare Flare and Sunburst on the same prompts and edit sequences; score identity preservation, text accuracy, latency, and review time.
2. Lock the creative brief. State the subject, composition, style, required text, background, aspect ratio, and what must not change.
3. Measure whole-request cost. Record input images, retries, quality setting, generated output, and human selection time instead of comparing only headline rates.
4. Keep provenance and review. Preserve prompts and source rights, inspect critical details, and account for C2PA metadata and watermarking in downstream workflows.
5. Verify the active route. Confirm the exact model ID, account access, limits, and current selling price before switching production traffic.

GPT Image 2.5 FAQ