Diskon 20% untuk waktu terbatas — paket bulanan mulai $4.99

Lihat paket
Third-party workspace · GPT Image 2.6 (Flare & Sunburst)

GPT Image 2.6 online: Flare vs Sunburst, credits, and a free try

This page is about using GPT Image 2.6 on this website. It is a third-party workspace at gptimage2.asia, not the official OpenAI product, and not ChatGPT. Official samples and model identifiers belong on the Images 2.6 model note. The model note’s call-to-action now points at GPT Image 2.6 on this site; GPT Image 2.5 remains available.

Generate with GPT Image 2.6

Guests get one free Flare 1K generation. The homepage FAQ says “one free still image.” Sign in to use Sunburst, 2K/4K, reference-image editing, GIF, and SVG. The full generator defaults to Sunburst.

Model

Resolution

Aspect ratio

GPT Image 2.6

Flare or Sunburst: which lane to start with

OpenAI describes Flare for faster iteration (it says Images 2.6 generates about 50% faster than before; at official quality=low, Flare can generate in as little as six seconds — that speed claim is OpenAI’s, not a test run on this site) and Sunburst for higher-detail work and steadier consistency across edits. Start with Flare at 1K while exploring. Switch to Sunburst, or to 2K / 4K, when the layout is close. If you only wanted a cheap draft, change the model to Flare before you press Generate — the full generator defaults to Sunburst. This site does not currently expose official API controls such as complexity, markup-guided edits, or a visible quality ladder on the generate page (checked 2026-11-12).

GPT Image 2.6 Flare

gpt-image-2.6-flare

Start with Flare at 1K when you are exploring composition, trying several wordings, or running many variations. It costs fewer credits than Sunburst at the same size.

  • Drafts, thumbnails and social posts
  • Testing many prompt variations quickly
  • Workloads where a faster, lower-cost first pass is enough

GPT Image 2.6 Sunburst

gpt-image-2.6-sunburst

Switch to Sunburst, or to 2K / 4K, when the layout is close: denser posters, small text you still need to proofread, product or face consistency, or a still you intend to ship. The full generator defaults to Sunburst.

  • Posters and layouts with exact text (always proofread)
  • Final marketing and product visuals
  • Multi-step edits that must keep identity and composition

Credits per image on this site

Failed generations are not charged: credits are deducted only after an image is returned successfully (re-checked 2026-11-12). The table below is driven by this site’s credit function and matches the published 2.6 hub rates. The pricing-page intro groups GPT Image 2.5 with 2.6 Flare at the same 2 / 3 / 4 ladder; this table lists GPT Image 2 at 1K but does not repeat a 2.5 row.

ModelResolutionCredits per image
GPT Image 2.6 Flare1K2
GPT Image 2.6 Flare2K / 4K3
GPT Image 2.6 Sunburst1K3
GPT Image 2.6 Sunburst2K / 4K4
GPT Image 21K1
  • Pack cards say “30 / 150 / 500 AI image generations,” some homepage copy says “30 credits,” and a Flare 1K still costs 2 credits. This page does not promise how many finished images a pack will produce. See the callout on Pricing.
  • GIF and SVG do not have a separate surcharge listed on the pricing page as of 2026-11-12 (not stated on the pricing page). Reverse-prompt analysis in the generator UI is described as costing 2 credits.
  • Batch generate keeps GPT Image 2.5 and 2.6 stills at 1K: Flare 2 credits each, Sunburst 3, according to the batch footer.
  • For Seedream 5.5’s separate 2 / 4 ladder, reference-image add-ons, and layer-split billing, use the pricing explainer and the Seedream 5.5 page.

See credit packs

A simple workflow on this site

  1. 1

    Open the generator with the lane you want

    Drafts: Generate with Flare. Finals: Generate with Sunburst. Community prompts on this site are still largely filed under GPT Image 2.5 prompts; if you reuse one, switch the picker to 2.6 yourself.

  2. 2

    Write or paste a prompt

    Say what the image is for, then the subject, framing, materials, and light. Put any exact words that must appear in quotation marks, say where they go, and ask for no extra text.

  3. 3

    Pick resolution

    Stay on 1K while you explore. 2K and 4K cost more credits and take longer.

  4. 4

    Optional: add a reference image (signed-in)

    Say what to change and list what must stay the same (face, product shape, logo, colors). Change one instruction per round. If something drifts, restate the must-stay list instead of rewriting the whole prompt. Upload only photos you have the right to use.

  5. 5

    Check the result

    Proofread spelling, identity, and whether the edit touched only what you asked for. Then download it, or reuse it as the next reference.

  6. 6

    Need a loop or a vector file?

    Follow the GIF and SVG guide. Guests cannot export SVG.

Prompting habits that transfer well

These are practical habits, not guarantees. They follow patterns in OpenAI’s image prompting guide and the existing copy on this hub.

State the outcome first

Then visible details. “Photorealistic” is a look you can ask for; lens names are hints, not a promise.

Give references a role

For several reference images, say which one supplies the subject, which one supplies the style, and which one supplies the background.

Put exact text in quotes

Quote every word that must appear, say where it goes and in what style, and ask for no extra text.

Edit with a must-stay list

Change only one thing and list what must not change. Name things to exclude, such as extra logos or watermarks.

Proofread required numbers and names

The model can render short text, but this site does not promise typo-free type.

Check the file you download

Verify spelling of required text, identity and product shape, and that the edit touched only what you asked for.

GPT Image 2.5 Prompts

What this site can and cannot do (GPT Image 2.6 path)

Can

  • Text-to-image and reference-image edits at 1K–4K without an OpenAI API key.
  • One guest try for a still image. This hub describes that try as one free Flare 1K generation. The homepage FAQ says “one free still image.” Animated GIF and SVG need a signed-in GPT Image 2.6 session; the guest trial cannot export SVG.
  • PNG, JPEG, GIF, and SVG on Flare and Sunburst, with the limits in the GIF and SVG guide.
  • Keep GPT Image 2.5 and GPT Image 2 available for comparison.

Cannot, or not offered here today

  • This is not ChatGPT and not the official API console. Official Images 2.6 is also available in ChatGPT, Codex, and the API.
  • No on-site markup or complexity controls were present on /generate as of 2026-11-12. Official quality tiers (low / medium / high / xhigh / max) are likewise not shown here.
  • SVG is text-to-image only, one at a time, not for reference edits, not usable as an edit input, not in batch generate, and not on the guest trial.
  • Seedream-only features (native 1.5K rung, layer split, Seedream’s transparent-PNG controls) are not part of the GPT Image 2.6 path.
  • This page does not promise a commercial license. Some paid packs mention one, but the live terms and privacy pages are still leftover templates. Treat licensing as not stated in the current policy until those pages are rewritten.

Official parameter docs live at OpenAI’s image generation guide. Do not expect to “circle an area” or set complexity here unless those controls are added later. OpenAI · Image generation guide

Frequently asked questions

What is GPT Image 2.6 on this site?

It is OpenAI’s Images 2.6 model family, offered here as two picker entries: gpt-image-2.6-flare and gpt-image-2.6-sunburst. Official background and samples belong on the Images 2.6 model note. This site is a third-party frontend.

How is it different from 2.5?

2.6 keeps the Flare / Sunburst split. OpenAI’s announcement emphasizes a faster Flare story and a higher-detail Sunburst lane, plus GIF and SVG in the official product. On this site, 2.5 remains selectable, and the credit ladder for Flare / Sunburst is described as matching 2.5. A fuller comparison is on GPT Image 2.6 vs 2.5.

Do I need an OpenAI API key?

No. Sign in and use this site’s credits. There is nothing to configure in an OpenAI dashboard for the on-site generator.

Which resolution should I choose?

Use 1K while exploring. Move to 2K or 4K when you are happy with the composition, knowing those sizes cost more credits.

Can I edit my own photos?

Yes, after you sign in: upload a reference and describe the change. Logged-in users can edit with both Flare and Sunburst. Upload only photos you have the right to use. This site’s help does not cover non-consensual face changes.

Are failed generations charged?

No. Live pricing and this hub both say credits are deducted only after a successful image is returned (re-checked 2026-11-12).

More about GPT Image models