🎨 Seedream 5.0 Pro — ByteDance's flagship image model: up to 2K, best-in-class typography & multi-reference editing!

Try It!
Logo
Image Models

Free GPT Image 2.5 Generator - Flare for Speed, Sunburst for Precision Editing

GPT Image 2.5 is OpenAI's image family released on 8 September 2026, and Dreamega gives you both API models in one place. Flare is OpenAI's default choice for most applications — the speed-optimised tier that runs at 50% lower latency than GPT Image 2, ideal for social content, product imagery, visual search and high-volume generation. Sunburst is the quality-optimised tier, built for premium workflows that need tighter control across edits, such as campaign creative and polished product shots. Both generate from text and edit existing images, support mask-guided inpainting, add the new xhigh and max quality tiers, and output up to 4K. No watermark, unlimited generations.

App versions cost fewer credits but may be less stable. For production work, choose the default (non-App) version for higher reliability.

GPT Image Model Series

GPT Image 2.5 Text to Image Gallery

Every image below was generated with gpt-image-2.5-flare through Dreamega at 2K, straight from the prompt shown. Flare is OpenAI's speed-optimised tier, running at 50% lower latency than GPT Image 2 with the sharper detail and stronger style adherence of the 2.5 generation.

Create with GPT Image 2.5
Neon Night Market

Neon Night Market

Multilingual neon signage rendered cleanly across a rain-washed alley, with photoreal reflections and depth.

“Cinematic wide shot of a rain-washed night market alley, dense hanging signage stacked overhead — glowing signs reading "HARBOUR NOODLE HOUSE", "汤面 · 24H", "ICE BAR", "LOTTERY 彩票" and a vertical banner "第七街夜市" — steam rising from a noodle stall, reflections pooling on the wet cobblestones, teal and magenta neon against warm tungsten, anamorphic flare, 35mm, shallow depth of field”

GPT Image 2.5 Image Edit Gallery

Each pair below was edited with gpt-image-2.5-sunburst, the precision tier built for tighter control across edits. One instruction changes one thing — composition, lighting and every untouched pixel stay as they were.

Edit with GPT Image 2.5
Product Colourway Swap - 1
Product Colourway Swap

Product Colourway Swap

01/05

What's GPT Image 2.5

OpenAI's September 2026 image models — Flare for fast everyday generation, Sunburst for precision editing

  1. 2API Models
  2. 50%Lower Latency
  3. 5Quality Tiers
  4. 4KMax Resolution

GPT Image 2.5 is OpenAI's image generation and editing family, launched on 8 September 2026 alongside ChatGPT Images 2.5. It ships as two API models: gpt-image-2.5-flare, the default choice for most applications, optimised for speed and running at 50% lower latency than GPT Image 2; and gpt-image-2.5-sunburst, optimised for quality and built for premium visual workflows that benefit from tighter control across edits. Both generate images from text and edit existing ones, support inpainting, and add two new quality tiers — xhigh and max — above the ceiling of earlier GPT Image models.

GPT Image 2.5's Powerful Features

What OpenAI shipped on 8 September 2026 with gpt-image-2.5-flare and gpt-image-2.5-sunburst

  1. 01

    Two Models, One Workflow

    Flare handles fast everyday generation while Sunburst handles precision editing. Switch between them on the same prompt without leaving the page or rebuilding your setup.

  2. 02

    50% Lower Latency

    Flare is the speed-optimised tier and runs at 50% lower latency than GPT Image 2, which makes it the practical choice for high-volume generation, visual search and rapid prototyping.

  3. 03

    Precision Localized Editing

    Sunburst is the precision tier, spending more time per image so campaign creative and polished product shots keep tighter control across every edit you apply.

  4. 04

    Sharper Detail, Stronger Style

    GPT Image 2.5 improves on sharpness of fine detail and on how closely output follows a requested style, compared with the GPT Image 2 generation it replaces.

  5. 05

    Mask-Based Inpainting

    Both models support inpainting on the image edits endpoint. Supply a mask with an alpha channel to point the model at the region you want redrawn, then describe the change in the prompt.

  6. 06

    Output Up to 4K

    Five quality tiers from low to max, square, landscape and portrait presets, plus custom sizes in multiples of 16 from 1:3 to 3:1 and up to 3840 pixels per edge.

  7. 07

    Dependable Text Rendering

    Headlines, product labels, signage and interface copy come out legible, so posters, packaging mockups and UI screens are usable without a manual typography pass.

  8. 08

    Consistent Across Edit Turns

    Repeated editing turns stay consistent with the reference image, so you can iterate on one asset through several instructions without the subject drifting.

GPT Image 2.5 YouTube Videos

Hands-on demonstrations, comparisons and walkthroughs of OpenAI's GPT Image 2.5 Flare and Sunburst models

  • Introducing GPT-Image-2.5 in the API - OpenAI
  • Introducing ChatGPT Images 2.5 - OpenAI
  • Sketch with ChatGPT Images 2.5 - OpenAI
  • Templates with ChatGPT Images 2.5 - OpenAI
  • I Tested GPT Image 2 vs New 2.5 (it's not even close) - Brendan Jowett
  • GPT-IMAGE 2.5: The New King of AI Images! - Theoretically Media

GPT Image 2.5 YouTube Videos

Hands-on demonstrations, comparisons and walkthroughs of OpenAI's GPT Image 2.5 Flare and Sunburst models

AI Pricing Plans

Flexible AI Pricing

Pay-as-you-go credits or subscription plans. No hidden fees, cancel anytime.

One Time supports crypto payment (BTC, USDT, ETH, 350+) · Monthly billing

Free

Try before you buy

Free
32points
  • Up to 3 videos
  • Up to 32 images
  • Multi-Model Support
  • Text to Video
  • Image to Video
  • Video to Video
Show more (+11)
  • Consistent Character
  • AI Animation Generator
  • Templates & Effects
  • AI Video Enhancers
  • Interactive Community
  • Faster Generation Speed
  • No-watermark Outputs
  • More Camera Movement
  • Private Video Visibility
  • Copy Protection
  • Priority Support

Pro

Popular

Elevate your AI experience

29.99USD / 1 Month
800
800 credits/mo
  • Up to 80 videos
  • Up to 800 images
  • 3 tasks(Parallel Tasks)
  • Multi-Model Support
  • Text to Video
  • Image to Video
Show more (+12)
  • Video to Video
  • Consistent Character
  • AI Animation Generator
  • Templates & Effects
  • AI Video Enhancers
  • Interactive Community
  • Faster Generation Speed
  • No-watermark Outputs
  • More Camera Movement
  • Private Video Visibility
  • Copy Protection
  • Priority Support

Lite

Start your AI journey

19.99USD / 1 Month
300 credits/mo
  • Up to 30 videos
  • Up to 300 images
  • 3 tasks(Parallel Tasks)
  • Multi-Model Support
  • Text to Video
  • Image to Video
Show more (+12)
  • Video to Video
  • Consistent Character
  • AI Animation Generator
  • Templates & Effects
  • AI Video Enhancers
  • Interactive Community
  • Faster Generation Speed
  • No-watermark Outputs
  • More Camera Movement
  • Private Video Visibility
  • Copy Protection
  • Priority Support
  • Card
  • Crypto accepted (BTC, USDT, ETH, 350+)
View Detailed Pricing

Frequently Asked Questions

Common questions about GPT Image 2.5 generation and editing

Still have questions?

[email protected]DiscordJoin our DiscordSubmit a Ticket
Flare is the fast tier and OpenAI's default choice for most applications: it is optimised for speed and runs at 50% lower latency than GPT Image 2, which suits creator and social content, product experiences, visual search, rapid prototyping and high-volume generation. Sunburst is optimised for quality — it spends longer per image and is built for premium visual workflows that benefit from tighter control across edits, such as production-ready campaign creative or polished product imagery. OpenAI's announcement credits Flare with higher image quality than GPT Image 2, while its API documentation describes Flare's quality as comparable to GPT Image 2 and Sunburst's as higher.
GPT Image 2.5 brings sharper detail, stronger style adherence and more control over edits than GPT Image 2, and it adds two quality tiers, xhigh and max, that earlier GPT Image models do not have. It also splits the family into two API models — gpt-image-2.5-flare and gpt-image-2.5-sunburst — so you can trade latency against editing precision. Flare runs at 50% lower latency than GPT Image 2, while Sunburst spends longer per image for extra fidelity.
OpenAI lists the same token rates for both GPT Image 2.5 models as for GPT Image 2: $5 per million text input tokens ($1.25 cached), $8 per million image input tokens ($2 cached) and $30 per million image output tokens. Text output is not billed because these models output images. Equal token rates do not mean equal cost per image, though — token consumption varies by model and quality setting. On Dreamega you pay in credits instead, and the exact figure for each quality and resolution combination is shown in the pricing table on this page.
OpenAI recommends 1024x1024 for square, 1536x1024 for landscape and 1024x1536 for portrait. Custom sizes are accepted as width by height in multiples of 16, with aspect ratios between 1:3 and 3:1 and a maximum of 3840 pixels on the longest edge. Quality can be set to low, medium, high, xhigh, max or auto, and providers expose those tiers up to 4K output.
Yes. Both Flare and Sunburst are available on the image edits endpoint as well as image generation, and both support inpainting. Upload your reference images and describe the change in plain language; for a targeted redraw you can supply a mask, which must carry an alpha channel marking the region to replace. Sunburst is the better pick when the edit has to preserve the rest of the frame exactly.
Start with Flare. It is OpenAI's stated default for most applications and its lower latency matters most when you are iterating or generating in volume. Move to Sunburst when precision is the point — final campaign artwork, packaging and product imagery, or multi-step edits where every untouched pixel has to stay untouched. On Dreamega you can switch between the two on the same prompt.

Privacy preferences

We use essential storage to run the service. Optional analytics, advertising and convenience features stay off until you choose to enable them. Privacy policy