On November 4, 2026, OpenAI released ChatGPT Images 2.6: the official announcement, API changelog, system card, ChatGPT release notes and X announcement (23:35 UTC) all landed the same day, with ChatGPT, Codex and the API going live simultaneously. It's the third GPT Image generation of 2026 — GPT Image 2 on April 21, 2.5 on September 8, and 2.6 on November 4; the gap between generations shrank from 20 weeks to 8.
On October 28 at 20:56 UTC, a commit to OpenAI's public OpenAPI repository added the markup parameter, described as "Supports GPT Image 2.5 and 2.6" — the first official appearance of "2.6". An October 31 commit reverted the description to "Supports GPT Image 2.5", but the name was already out. OrcaRouter traced the leak to commit aa88004. On launch day itself, LMArena's T2I and editing boards gained 2.6 entries (marked Preliminary); fal shipped 2.6 endpoints the same night; Recraft Studio and Runware added it the next day; on Nov 7 Artificial Analysis listed both 2.6 models on its boards.
2.6 ships as a Flare + Sunburst pair: Flare for fast iteration, as fast as ~6s at low quality with improved layout quality; Sunburst for detailed visual work, higher quality at medium/high and stronger multi-round editing consistency. Note the dual-model split is not new to 2.6 — 2.5 already shipped as a Sunburst + Flare pair in September. Token prices are identical for both, and token use at the same settings measures the same, so the choice trades time for detail, not money. OpenAI hasn't disclosed which variant ChatGPT calls.
The official changelog is clear: 2.6 adds three things — animated GIF generation (looping GIFs, adjustable speed, frame-by-frame editing), SVG export (API-only), and the complexity planning parameter (controls how much planning happens before rendering, default high, independent of quality). The easy confusion is markup editing — it went to supported 2.5 models at the same time, so it's not a 2.6 exclusive; markup was already available in ChatGPT, and this release brings it to the API. Another confusing number is the GIF frame count: launch-day docs said 2–64 frames; after Nov 7 the docs read 2–16 frames (default 9); 64 is an old docs number.
First, price: token rates are identical to 2.5 (text $5, image input $8, image output $30 per 1M tokens), but per-image token use at low/medium quality is 4.2x/3.3x higher — low $0.02469 (2.5: $0.00588), medium $0.04347 (2.5: $0.01317). The official guide warns explicitly: "Equal token rates don't mean equal cost per image." Second, speed: the announcement says "50% faster" but names no comparison baseline; third-party tests show Flare genuinely faster (QuantSlant median 8.0s vs 2.5 Sunburst's 17.9s) while Sunburst 2.6 is slower (median 26.3s). "50%" cannot be written as a general speed-up for both models.
Six days after launch, on Nov 10, GPT Image 2 was marked deprecated, with removal set for 2027-02-10 and GPT Image 2.6 as the designated replacement. On Dec 1, gpt-image-1-mini, gpt-image-1.5 and chatgpt-image-latest retire (sources disagree on the designated replacement, so this site records only the retirement fact). On independent scores: 2.6 Sunburst is already #1 on both AA boards (T2I 1261, editing 1210), while 2.5 Sunburst still leads both LMArena boards — 2.6 has only a few thousand votes and is marked Preliminary, so the two boards don't contradict each other; it's a sample-size difference. What to watch next: follow-up tests of SVG reliability (CellCog says it re-runs weekly), the full ChatGPT rollout of GIF generation, and any official word on the free tier's quota. This site will keep tracking it.
Reference: OpenAI — Introducing Images 2.6 (rewritten and organized by ModelBrief; facts per the official source)