GPT Image 2.6 is the image model family OpenAI released on November 3–4, 2026, shipping simultaneously on ChatGPT, Codex, and the API. It adds GIF generation, SVG output, and photo markup; the API keeps the Flare (speed) and Sunburst (quality) variants; official pricing remains unpublished as of November 5, 2026. For scale: “gpt image 2.5” draws roughly 70K searches a month at a difficulty of just 24, and the equivalent 2.6 queries are expected to climb fast.
Key takeaways
- Released November 3–4, 2026, available simultaneously on ChatGPT, Codex, and the API (launch coverage).
- Three new capabilities: GIF generation (ChatGPT and API), photo markup, and SVG output (API only).
- Two API variants: Flare for faster generation, Sunburst for higher quality and cross-turn editing consistency.
- A new configurable reasoning effort trades generation speed against output refinement.
- As of 2026-11-05, official API pricing is unpublished; this article lists only confirmed facts and keeps unknowns in their own section — no number guessing.
- OpenAI also published a System Card with dedicated safeguards for GIF generation.
What is GPT Image 2.6?
GPT Image 2.6 is the latest version of OpenAI’s image generation and editing model family, released November 3–4, 2026. It generates or edits images from text prompts and reference images, and for the first time in this line it supports animated GIFs. The model is available to consumers (ChatGPT), coding contexts (Codex), and developers (API). Compared with Images 2.5, the main gains are editing consistency, detail retention, and understanding of sketches and photo markup.
Specs at a glance
| Spec | Current status |
|---|---|
| Developer | OpenAI |
| Model type | Image generation + editing family (incl. GIF) |
| Input modalities | Text prompt, reference image, photo markup |
| Output formats | Bitmap; SVG (API only); GIF (ChatGPT + API) |
| API variants | GPT-Image-2.6 Flare / GPT-Image-2.6 Sunburst |
| Reasoning effort | Configurable in both the Chat model picker and the API |
| Release date | November 3–4, 2026 |
| API pricing | Not yet published (as of 2026-11-05) |
| Availability | ChatGPT, Codex, API; third-party platforms rolling in |
| Open weights | Not announced |
Entries marked “not yet published” reflect the official state of information as of 2026-11-05 — they are not permanent facts. Check the latest OpenAI pages before citing.
What it can do: five typical scenarios
- Animated GIFs: generate motion from a prompt, or animate a static image — the first native GIF support in an OpenAI image model, open on both ChatGPT and the API (Decrypt).
- Markup feedback: draw, circle, and annotate on an image so the model edits by intent — useful for design review and asset rework.
- Vector workflows: SVG output via the API covers logo drafts and scalable icons that bitmap models can’t deliver directly.
- Multi-turn refinement: preservation-style edits to confirmed regions, with better consistency than 2.5 (vendor claim; independent benchmarks pending).
- Speed/quality switching: draft at low reasoning effort and finalize at high effort inside one integration — no need to swap models.
How it works: what changed this generation
OpenAI has not published architecture details for Images 2.6 — parameter counts, training data, and internals remain undisclosed, consistent with the 2.5 launch. So this article doesn’t argue about “stronger internals”, only about observable workflow differences: if 2.5 was about “don’t break the original while editing”, 2.6 is more like an art director who reads your notes — it treats sketches as editing instructions, keeps confirmed regions stable, and extends the same ability to motion.
Flare vs. Sunburst. Flare targets latency-sensitive flows: chat interfaces, live previews, batch first drafts. Sunburst spends longer generation time on higher quality and cross-turn consistency — right for marketing assets and product images where “the first image is only the starting point”. The naming and split carry over from Images 2.5 (launch documentation). A full numeric comparison between the two is unpublished; pick by workflow priority rather than waiting for benchmarks.
Reasoning effort. 2.6 lets you set the reasoning investment before generation: low effort for fast iteration, high effort for detail and instruction following. The option appears in both the Chat model picker and the API, which means a production pipeline can expose “draft/final” as one user-visible dial instead of maintaining two model integrations.
How to get it: three access channels
| Channel | Status | Notes |
|---|---|---|
| ChatGPT (consumer) | Live | GIF and photo markup in-app |
| Codex | Live | Open with the launch |
| Official API | Live | Flare + Sunburst; SVG output supported |
| Third-party/aggregators | Rolling in | Check each platform’s model catalog |
For developers, the real question is usually not “is the model available” but “which channel”: the official API requires the OpenAI account system and is region-restricted; aggregators typically offer the same models behind a single API key with unified billing, fitting teams that run models in parallel. If you’re reading this on API Hop, check the model catalog for GPT Image 2.6 availability.
Pricing: known and unknown
As of November 5, 2026, OpenAI has not published GPT Image 2.6 API pricing. No guessing — references only:
| Item | Status |
|---|---|
| GPT Image 2.6 API rates (text/image input, image output) | Not yet published |
| GPT Image 2.5 reference | Official per-image rates were never fully public; whether 2.6 reuses, adjusts, or restructures billing is unknown |
| GIF billing | No documentation — per output or per frame is unclear |
| Reasoning-effort billing | No documentation — whether higher effort costs more is unknown |
Once official numbers land, this table will be updated and the updated date refreshed. Any third-party “confirmed price” circulating before OpenAI publishes its own page should be treated as unverified.
Compared with the previous generation
| Dimension | GPT Image 2.6 | GPT Image 2.5 | Notes |
|---|---|---|---|
| GIF generation | ✅ | ❌ | Headline new feature |
| SVG output (API) | ✅ | ❌ | New |
| Photo markup understanding | ✅ | Weaker | Vendor describes it as improved |
| API variants | Flare + Sunburst | Flare + Sunburst | Naming carried over |
| Configurable reasoning effort | ✅ | ❌ | New |
| Pricing | Unpublished | Partially confirmed | Official sources rule |
A broader comparison against current Google and Midjourney image models is planned as a separate article — this one stays focused on the 2.6 launch itself.
What we still don’t know
- API pricing and billing structure, including how GIFs and reasoning effort are metered.
- Rate limits and regional availability for the two API variants.
- Maximum resolution and aspect ratio support — not mentioned in launch materials.
- Whether Flare and Sunburst are priced differently, plus consistency and detail results under independent benchmarks.
What to watch next
- Official pricing: on the day it lands, “GPT Image 2.6 pricing” searches will spike (the 2.5 equivalent draws ~70K monthly searches at difficulty 24); this article will be updated then.
- Aggregator rollout speed: which platforms list the model in week one is a practical variable for channel choice.
- GIF edge cases: early community tests will show how mark-up-driven animation holds up in complex scenes.
FAQ
Is GPT Image 2.6 free?
Inside ChatGPT it uses your plan’s existing image quota; the API is pay-as-you-go, but official per-image rates are unpublished as of 2026-11-05. Follow the latest OpenAI pricing page — this article will update once official numbers appear.
How much does GPT Image 2.6 cost?
Official API pricing is unpublished (as of 2026-11-05). The previous generation’s rates were only partially confirmed either, so treat any specific price circulating before OpenAI’s own page as unverified; we’ll add the full price table once confirmed.
When was GPT Image 2.6 released?
November 3–4, 2026, simultaneously on ChatGPT, Codex, and the API, per OpenAI’s launch announcements and the Images 2.6 System Card.
Can GPT Image 2.6 generate GIFs?
Yes. GIF generation is the headline feature of this generation, supported in both the ChatGPT app and the API, with dedicated safety measures.
Flare or Sunburst — which should I pick?
Pick Flare for low latency and fast output; Sunburst for higher quality and cross-turn editing consistency. The naming carries over from Images 2.5, there’s no full numeric comparison yet, so choose by workflow priority.
Does GPT Image 2.6 have an API?
Yes. Both variants (GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst) entered the API with the launch and support SVG output. Third-party platforms are rolling in.
What changed from 2.5 to 2.6?
The main additions are GIF generation, SVG output, and configurable reasoning effort, plus improved photo markup and sketch understanding. Whether the billing structure changed is unclear — see the comparison table above.
Can I use GPT Image 2.6 commercially?
Follow the current terms of whichever entry point you use (ChatGPT plan or API terms). No 2.6-specific commercial restrictions were announced at launch; verify the latest official terms before large-scale commercial deployment.
Information verified 2026-11-05: OpenAI Images 2.6 System Card · launch coverage · Decrypt · AI Matter. Pricing will be updated along with the updated date once published.