ReleaseAPI / Product · ActionableTG: @neyr0graph (NeuroGraph)→ original

OpenAI Releases GPT Image 2.6: GIFs, SVG Export and 50% Faster Generation

OpenAI has released GPT Image 2.6 in ChatGPT, Codex and the API. The company claims generation is 50% faster, with more consistent results across edits. You can now create looping GIFs and export images as SVG. There are two models: the fast Flare for drafts and Sunburst for detailed edits.

AI-processed from TG: @neyr0graph (NeuroGraph); edited by Hamidun News
OpenAI Releases GPT Image 2.6: GIFs, SVG Export and 50% Faster Generation
Source: TG: @neyr0graph (NeuroGraph). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @neyr0graph (NeuroGraph) ↗
Editor-in-Chief
Key metric #1At Low quality, Flare can deliver an image in 6 seconds
Key metric #2Edits using circles and arrows are now available in the API
Key metric #3Low costs $0.02469 per 1024×1024 image, Max costs $0.21072

Everyone was waiting for a new language model from OpenAI, but the company unexpectedly updated its image generation instead. GPT Image 2.6 is out, and for those building products with images, it is arguably more useful than yet another benchmark bump.

Two Models for Different Tasks

The update is available in ChatGPT, Codex and the API. OpenAI claims generation is 50% faster and that edits are more consistent with one another. There are two models.

Flare is for quickly exploring ideas: first drafts of a poster, an illustration or a product card. At Low quality, generation can take about six seconds. This is a claimed best-case scenario, not a guarantee for every request.

Sunburst is built for detail and for preserving consistency during editing. For such tasks OpenAI recommends Medium or High. A typical e-commerce example: replacing a product's background while keeping the shape, material and texture of the item itself. This is exactly where models used to love "inventing" extra details.

GIFs, SVG and Markup-Based Edits

The key new capabilities are as follows:

  • Looping GIFs from text or an existing image, with adjustable speed and editing of individual frames. Suitable for mascots, reactions, short object animations and storyboards.
  • SVG export for websites, apps and print. Vector graphics scale without pixelation, which means logos and icons can be produced without manual drawing.
  • Markup-based edits, which were already available in ChatGPT, are now in the API as well. Using circles, arrows and notes, you can show exactly where a change is needed.
  • In ChatGPT you can save images of yourself, your pets and characters and reuse them without uploading again.

Two separate parameters have also appeared. Quality controls visual detail. Complexity determines how much planning the model spends on a task. For most edits OpenAI recommends Low complexity, and for scenes with many objects and constraints, Medium or High. These settings should not be confused.

How Much It Costs

According to the official calculator, a 1024×1024 output image via the API costs the following:

  • Low: $0.02469 (about 2.02 RUB)
  • Medium: $0.04347 (about 3.55 RUB)
  • High: $0.05268 (about 4.30 RUB)
  • XHigh: $0.09366 (about 7.65 RUB)
  • Max: $0.21072 (about 17.21 RUB)

The prompt text and uploaded references are billed separately. SVG export may also include a charge for the model that builds the vector. So these figures are not the full price of any given request or finished SVG.

What to Do in Practice

Product teams and developers should rebuild their pipeline in two steps: drafts on Flare at Low, final output and edits on Sunburst at Medium or High. This keeps the cost of ideation minimal, and the expensive modes are used only where precision is needed.

It is worth testing markup-based edits in the API on your own real cases: if they hold the product's shape and texture, manual retouching will drop noticeably. Be more careful with SVG: calculate the full cost on real tasks rather than going by the raster price, and check vector quality on complex illustrations.

The models are also available through third-party services such as fal.ai and the Import Model node in Weavy AI, so you can try them without direct integration. The claimed 50% speedup and six seconds are best treated as a benchmark for now, and measured yourself.

Story

Key Questions Answered

What are the key details on Two Models for Different Tasks?

The update is available in ChatGPT, Codex and the API. OpenAI claims generation is 50% faster and that edits are more consistent with one another. There are two models. Flare is for quickly exploring ideas: first drafts of a poster, an illustration or a product card. At Low quality, generation can take about six seconds. This is a claimed best-case scenario, not a guarantee for every request. Sunburst is built for detail and for preserving consistency during editing.

What are the key details on GIFs, SVG and Markup-Based Edits?

The key new capabilities are as follows: Looping GIFs from text or an existing image, with adjustable speed and editing of individual frames. Suitable for mascots, reactions, short object animations and storyboards. SVG export for websites, apps and print. Vector graphics scale without pixelation, which means logos and icons can be produced without manual drawing. Markup-based edits, which were already available in ChatGPT, are now in the API as well.

How Much It Costs?

According to the official calculator, a 1024×1024 output image via the API costs the following: Low: $0.02469 (about 2.02 RUB) Medium: $0.04347 (about 3.55 RUB) High: $0.05268 (about 4.30 RUB) XHigh: $0.09366 (about 7.65 RUB) Max: $0.21072 (about 17.21 RUB) The prompt text and uploaded references are billed separately. SVG export may also include a charge for the model that builds the vector. So these figures are not the full price of any given request or finished SVG.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Hands-On AI Engineering · Hamidun Academy

Build & Ship Production AI Agents Without Boilerplate Code

Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.

Start 72h Free Trial →Vibe-Coding Intensive
What do you think?
Loading comments…