AI SINGLE SOURCE

OpenAI rolls out GPT-Image-2.6 as Flare and Sunburst across ChatGPT, Codex and API

First reported 05 Nov, 07:03 UTCSource: MarkTechPostLast editorial activity 05 Nov, 07:03 UTCWrite a comment ↓

Abstract 3D render visualizing artificial intelligence and neural networks in digital form. — Illustrative photo.Photo by Google DeepMind on PexelsOpenAI logo shown for identification only; no affiliation with or endorsement of WORLDTECH is implied.

OpenAI has released Images 2.6 across ChatGPT, Codex and its API, with GPT-Image-2.6 arriving as two models, Flare and Sunburst, according to a report by MarkTechPost. The update adds GIF output, SVG export, markup-guided edits and a new complexity setting.

Two models aimed at different work

MarkTechPost reports that Flare is intended for quick iteration and interactive workflows, whereas Sunburst is pitched at detailed compositions and editing. OpenAI states that Flare can return an image in as little as 6 seconds when quality is set to low.

Sunburst is positioned for higher quality and steadier consistency across successive edits when quality is set to medium or high.

New creative tools inside ChatGPT

ChatGPT users can now convert a written prompt or a photograph into an animated GIF, change the speed of the animation, edit single frames and share the output. Reference images of people, pets and characters can also be saved, labelled and called up again later without being uploaded anew.

Complexity setting and markup edits

A new complexity field determines how much planning the model does before rendering. The image generation guide lists three levels, low, medium and high, and both models default to high when the field is left out. According to the guide, low suits narrow changes such as swapping a background or a sign, while high handles dense text, intricate layouts and the relationships between objects. OpenAI cautions that higher complexity can increase latency and bring in unintended alterations. This setting is distinct from quality, which governs rendering detail and fidelity.

Markup lets users indicate the precise region they want changed. The launch post says markup support is coming to the API.

Usage figures and speed claims

OpenAI says more than 4B images are now generated per week across ChatGPT and the API, up from more than 3B per week eight weeks earlier at the Images 2.5 launch. That works out to over 500M images per day on average. OpenAI also claims Images 2.6 produces images 50% faster than the previous release. These are company statements as relayed by MarkTechPost.

Open questions

Will markup-guided edits reach the API on a stated date?
Will benchmark scores or per-image cost figures be published for GPT-Image-2.6?

Key facts

OpenAI has shipped Images 2.6 across ChatGPT, Codex and its API, according to MarkTechPost.

The core developer release is GPT-Image-2.6, offered as two API models named Flare and Sunburst.

Flare can produce an image in as little as 6 seconds when quality is set to low.

OpenAI states that Images 2.6 generates images 50% faster than the previous release.

OpenAI reports more than 4B images generated per week across ChatGPT and the API.