Introduction

On November 4, 2026, OpenAI shipped Images 2.6 into ChatGPT, Codex, and the Images API. The headline is speed plus edit control: generation is claimed ~50% faster, API users pick between GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst, and makers finally get animated GIF output, SVG exports, and markup-guided edits in the API — not just ChatGPT.

This is the GA follow-through after the late-October API preview chatter. Named models, global ChatGPT availability, and day-0 fal endpoints make it a real run-today ship.

What shipped

Concrete facts from OpenAI’s blog, API changelog (Nov 4), and model cards:

  • Product: Images 2.6 across ChatGPT, Codex, and the API
  • Vendor: OpenAI (closed / API)
  • Modality: text-to-image and image edit, including animated GIF and SVG output
  • API models: gpt-image-2.6-flare (fast iteration) and gpt-image-2.6-sunburst (higher detail / edit consistency at medium–high quality)
  • Speed claim: Images 2.6 ~50% faster; Flare can hit ~6 seconds when quality is set to low
  • New knobs: separate quality (visual fidelity) and complexity (how much planning the model does); markup overlays for scoped edits on /v1/images/edits
  • Endpoints: v1/images/generations and v1/images/edits
  • Availability: ChatGPT + Codex globally on all subscription tiers (desktop, mobile, web)

Flare vs Sunburst

Think of the pair as two dials on the same stack:

  • Flare — everyday generation and quick edit loops. Best when you want layout-friendly posters and product shots without waiting. Low quality can land around six seconds.
  • Sunburst — higher-fidelity generation and tighter character / product consistency across edit rounds when quality is medium or high.

OpenAI’s guidance for most edits: keep complexity low for speed, then raise quality only as far as the job needs. Bump complexity for crowded scenes with many constraints.

Where to run it today

  1. ChatGPT / Codex — Images 2.6 is on for every subscription tier; markup editing that was ChatGPT-only now also lands in the API
  2. OpenAI Images API — set model to gpt-image-2.6-flare or gpt-image-2.6-sunburst on generations and edits
  3. fal — day-0 catalog endpoints for Flare and Sunburst text-to-image and edit (markup + GIF called out on the fal drop)

Example edit sketch (markup + prompt; do not combine markup with a classic mask):

curl https://api.openai.com/v1/images/edits \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -F model="gpt-image-2.6-flare" \
  -F prompt="Recolor the circled vase cobalt blue" \
  -F image[]="@product.png" \
  -F markup="@markup.png" \
  -F quality="medium" \
  -F complexity="low"

Flare token card (OpenAI docs): text input $5 / 1M, cached text $1.25 / 1M; image input $8 / 1M, cached image $2 / 1M, image output $30 / 1M. Quality strongly changes how many image tokens you burn.

Why it matters

The Oct preview already teased markup and SVG. GA locks the story makers actually ship with: a fast Flare lane, a precise Sunburst lane, motion via GIF for social and mascot loops, and vector SVG for web/print handoff — all under named API IDs you can pin in production.

For ArtRealm readers who live in open Comfy graphs, this is a closed stack. For teams already on ChatGPT Images or the Images API, it is the iteration upgrade to wire today.

Caveats

  • Closed weights — no local Comfy twin of Flare/Sunburst
  • GIF / SVG / markup behavior is model- and endpoint-specific; verify output format on your client
  • Cost scales with quality and image tokens; “six seconds” is the low-quality Flare path, not every call
  • ArtRealmAI Gen does not host OpenAI Images 2.6 — run ChatGPT, the API, or fal
  • Safety / system-card evaluations for 2.6 are separate reading if you ship client-facing brands

Original Source

https://openai.com/index/introducing-images-2-6/

Conclusion

Images 2.6 is the GA ship: Flare when you need the draft now, Sunburst when the edit round has to hold, plus GIF, SVG, and markup so the API matches how people already work in ChatGPT. Pin the model IDs, pick your quality/complexity pair, and iterate.

—Aurelia ♡