GPT Image 2.5 — Release, API & Pricing

Release timeline

What's New in GPT Image 2.5 — Release Date & Timeline

The GPT Image 2.5 release date is September 8, 2026. OpenAI announced it on the consumer side as ChatGPT Images 2.5 — you will also see it written as images 2.5, gpt images 2.5 or chatgpt images 2.5 — and shipped the matching API models the same day. The headline changes are speed and edit stability rather than a wholesale change in visual style: generation latency in the ChatGPT client is roughly half what the 2.0 generation delivered, and iterative editing holds character, layout and lighting far more consistently across several rounds of instructions.

1. 2026-09-08

   GPT Image 2.5 announced

   OpenAI publishes the ChatGPT Images 2.5 release and opens the new image models to API users the same day.

2. Consumer rollout

   ChatGPT Image 2.5 in the app

   Image generation inside ChatGPT switches to the 2.5 generation, with generation latency reduced by about 50% compared with the 2.0 experience.

3. API availability

   Two model IDs go live

   gpt-image-2.5-flare ships as the default general-purpose model; gpt-image-2.5-sunburst ships alongside it for higher-precision editing work.

4. Ongoing

   Pricing and limits still landing

   Per-image and per-token rates for gpt-image-2.5 are not fully published yet. Always confirm current numbers on platform.openai.com before you budget.

GPT Image 2.5 API

API Access — Endpoints & Model IDs

GPT Image 2.5 API access runs through the standard OpenAI API image surface. If you already call the GPT API or the Chat GPT API from your app, nothing structural changes: you authenticate with the same OpenAI API key, POST to the images endpoints, and select the generation by model string. The two model IDs are gpt-image-2.5-flare and gpt-image-2.5-sunburst.

curl https://api.openai.com/v1/images/generations \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.5-flare",
    "prompt": "a ceramic coffee cup on a linen tablecloth, morning window light",
    "size": "1024x1024"
  }'

Request and response fields, supported sizes and rate limits are defined by OpenAI and can change without notice — check the official API reference at platform.openai.com before shipping.

Model variants

Flare vs Sunburst — Which gpt-image-2.5 Model to Use

The practical decision with gpt-image-2.5 is which of the two model IDs you pass. Flare is the default and is tuned for throughput: it is the one to reach for when you are generating fresh images, running batch jobs, or letting an end user iterate interactively and waiting is expensive. GPT Image 2.5 sunburst trades speed for precision on edits — it holds fine detail, small text and masked-region boundaries more faithfully, and takes longer to return. Many teams end up using both: flare for drafting, sunburst for the final pass.

Both variants sit behind the same endpoints — switching between them is a model-string change, not a rewrite.

GPT Image 2.5 pricing

GPT Image 2.5 Pricing — What Is and Isn't Confirmed

This is the honest answer on gpt image 2.5 pricing: OpenAI has not fully published the per-token or per-image rates for gpt-image-2.5-flare and gpt-image-2.5-sunburst. Anything you see quoted as an exact number for these models right now should be treated as unverified. We will not invent figures here.

Need predictable image costs today? xImagineai sells one-time credit packs with no OpenAI account and no per-image billing surprises.

Competitive landscape

GPT Image 2.5 vs Nano Banana, Seedance, Gemini 2.5 Flash Image & Midjourney

GPT Image 2.5 does not land in an empty field. Google's Gemini 2.5 Flash Image — widely known by its nickname Nano Banana — is the closest competitor on conversational editing and speed. Seedance comes from a different direction: it is a video-generation model, so it competes for the same creative budget rather than the same output type. Midjourney remains the reference point for stylised aesthetic quality, with a workflow built around its own app rather than an API-first developer surface. Here is how the 2.5 generation lines up.

Competitor capabilities and pricing move quickly — verify each vendor's current documentation before making a platform decision.

Related Pages

GPT Image 2.5 is a separate model generation from GPT Image 2 — different model IDs, different behaviour. If you landed here looking for the previous release, start with the GPT Image 2 guide.

• GPT Image 2 (gpt-image-1) — previous generation guide — Features, prompt optimizer, published API pricing and alternatives for the 2.0 generation.
• Grok Imagine 2.0 vs GPT Image 2 — Head-to-head comparison of the two models that topped the image arena.
• Grok Imagine vs Midjourney — How an API-first image model compares with Midjourney's app-driven workflow.
• xImagineai pricing — One-time credit packs for image and video generation, no subscription.