GPT Image 2.5 introduces two specialized image generation models — Flare for high-speed production and Sunburst for precision editing — along with a Sketch drawing tool, comment-based image editing, and output up to 4K resolution.
OpenAI today released GPT Image 2.5, its latest image generation platform powering over 3 billion images per week. The update splits image creation into two purpose-built models: GPT Image 2.5 Flare for speed-first workflows and GPT Image 2.5 Sunburst for precision-first creative production, while introducing new creation tools including Sketch, Templates, and comment-based editing.
SEPTEMBER 8, 2026 – OpenAI today announced the launch of GPT Image 2.5, its most advanced image generation platform to date. The release introduces a dual-model architecture that gives developers and creators the flexibility to choose between speed and precision depending on their workflow.
Available immediately through the OpenAI API and ChatGPT, GPT Image 2.5 replaces the single-model approach of its predecessor with two specialized models: GPT Image 2.5 Flare for high-throughput, cost-efficient generation and GPT Image 2.5 Sunburst for maximum editing control and production-grade creative quality.
Background: Why a Dual-Model Approach
AI image generation has matured beyond experimentation into daily creative production. Teams across marketing, e-commerce, publishing, and design now need different things from the same technology: some workflows demand maximum speed for rapid iteration, while others require maximum precision for brand-critical deliverables.
Previous single-model architectures forced users to compromise — accept slower generation for quality, or sacrifice editing fidelity for speed. GPT Image 2.5 addresses this fundamental tension by offering two models at the same token price, letting teams route each request to the right model for the job.
GPT Image 2.5 Flare: Speed-First Image Generation
GPT Image 2.5 Flare is OpenAI's speed-optimized "small" model and the default recommended choice for most API applications. It delivers significantly lower latency and 2–4× faster throughput than GPT Image 2, with comparable or better image quality.
Early partner Manus confirmed real-world 2–4× speed improvements along with better transparent-background output quality.
Key capabilities include:
2–4× faster throughput compared to GPT Image 2, validated by early partners
Default API recommendation for most image generation applications
Quality tiers from low to max with flexible resolution up to 4K (3840×2160)
Transparent background support with improved alpha-channel output
Reference-guided creation for consistent visual style across batches
Same token pricing as Sunburst — no speed premium
GPT Image 2.5 Flare is best suited for social media content, e-commerce product images, marketing asset generation, rapid prototyping, visual search, and any workflow where speed and volume matter.
GPT Image 2.5 Sunburst: Precision-First Creative Editing
GPT Image 2.5 Sunburst is OpenAI's precision-optimized "base" model, built for production-ready creative work that demands the tightest editing control and highest image quality.
Early partner Higgsfield AI noted that the model "understands what not to change" — preserving unedited regions with greater fidelity across successive rounds of editing.
Key capabilities include:
Highest image quality in the GPT Image family, surpassing GPT Image 2
Tightest multi-turn editing control — changes only what is asked while preserving everything else
Best subject preservation — faces, products, and brand elements remain consistent across edits
Complex composition handling — accurate editing even in busy, multi-subject scenes
Up to 4K resolution output with five quality tiers (low, medium, high, xhigh, max)
Lowest unsafe generation rate at 1.09% (vs. 1.41% for Flare, 1.64% for GPT Image 2)
GPT Image 2.5 Sunburst is best suited for campaign creative, polished product photography, brand collateral, magazine-quality imagery, multi-round refinement workflows, and design tool integration where precision and brand accuracy are non-negotiable.
New Creation Tools in ChatGPT
GPT Image 2.5 introduces three new creative tools available across all ChatGPT tiers:
Sketch Tool — Type @Sketch in ChatGPT to draw rough layouts and concepts directly in the chat. The model transforms sketches into polished, finished images. As OpenAI noted, "Some ideas are easier to draw than describe."
Comment-Based Editing — Place comments directly on specific spots in an image (e.g., "make this darker", "change this to blue") for precise, localized edits without re-describing the entire scene in text.
Templates — Pre-built templates for posters, merchandise, flyers, and product photos. Users add their own text, design elements, or style on top, lowering the barrier to creating from scratch.
Availability
GPT Image 2.5 is now available through the OpenAI API and ChatGPT across all tiers (Free, Plus, Pro, Work). Developers can access Flare and Sunburst via the Images API (v1/images/generations, v1/images/edits) and the Responses API.
Frequently Asked Questions about GPT Image 2.5
Q: What is GPT Image 2.5?
A: GPT Image 2.5 is OpenAI's latest image generation platform, released September 8, 2026. It introduces two specialized models — Flare for speed and Sunburst for precision — along with new tools including Sketch, Templates, and comment-based editing.
Q: What is the difference between GPT Image 2.5 Flare and GPT Image 2.5 Sunburst?
A: GPT Image 2.5 Flare is the speed-optimized model delivering 2–4× faster throughput, ideal for batch generation, marketing, and e-commerce. GPT Image 2.5 Sunburst is the precision-optimized model with the tightest editing control, best for campaign creative, product photography, and multi-round refinement.
Q: How much does GPT Image 2.5 cost?
A: Flare and Sunburst share the same token pricing: $5/1M text input, $8/1M image input, $30/1M image output. Per-image costs range from ~$0.006 (low quality, 1K) to ~$0.211 (max quality, 1K). The high tier is approximately 4× cheaper than the equivalent in GPT Image 2.
Q: Is GPT Image 2.5 backward-compatible with GPT Image 2?
A: Yes. All API endpoints are unchanged. To migrate, simply swap the model ID from gpt-image-2 to gpt-image-2.5-flare or gpt-image-2.5-sunburst. GPT Image 2 remains available.
Q: What is the Sketch tool?
A: Sketch is a new drawing tool in ChatGPT. Type @Sketch to draw rough layouts and concepts, then let the model transform them into polished images. It bridges the gap between visual ideas and text-only prompting.
Q: Does GPT Image 2.5 support transparent backgrounds?
A: Yes. Both Flare and Sunburst support true alpha-channel transparent background output in PNG format, now generally available (previously in preview).
Q: What resolution does GPT Image 2.5 support?
A: Both models support resolutions up to 4K (3840×2160), with custom sizes in multiples of 16 and aspect ratios from 1:3 to 3:1.
Q: How does GPT Image 2.5 handle content safety?
A: GPT Image 2.5 uses a multi-layered safety system including prompt filtering, input classifiers, joint edit analysis, and output classifiers. All generated images include C2PA metadata and SynthID invisible watermarks for content provenance verification.
About GPT Image 2.5
GPT Image 2.5 is OpenAI's next-generation image generation platform, powering over 3 billion images per week across ChatGPT and the OpenAI API. The platform splits image creation into two purpose-built models — Flare for high-speed production and Sunburst for precision editing — and introduces new creative tools including Sketch, Templates, and comment-based editing to make AI image creation more intuitive and production-ready.
Top comments (0)