Blog · 2026-11-05

GPT Image 2.6 vs Grok Imagine Image 2.0: More Formats or Flat Tiers

Model pages: GPT Image 2.6 Flare API · GPT Image 2.6 Sunburst API · Grok Imagine Image 2.0 API — pricing, specs and the playground

Pick GPT Image 2.6 when the job needs animated GIF, SVG, a transparent background, 4K or more than three reference images. Pick Grok Imagine Image 2.0 when you want a fixed price per image, from $0.03, for 1K or 2K work with up to three references.

The deciding fact is billing. GPT Image 2.6 is metered per token on apimodels.app, so quality, size, format, references and prompt length all move the price. Grok Imagine Image 2.0 has four flat tiers: 1K low $0.03, 2K low $0.04, 1K medium $0.045 and 2K medium $0.06, with references included. In our 5 November 2026 tests, a 1K low GPT Image 2.6 request came to about $0.0077 and a 2K medium one to $0.0516. Both are below the matching Grok tier, though they are measured examples, not a price list.

GPT Image 2.6 is OpenAI's model and comes in two tiers: Flare, the default, and Sunburst, the highest-fidelity tier. Both take the same parameters and bill at the same token prices. Grok Imagine Image 2.0 is a single model from xAI.

GPT Image 2.6 vs Grok Imagine Image 2.0 at a glance

Pricing — GPT Image 2.6 charges per token: image output $30 per 1M, image input $8, text input $5 and text output $10. Each request is rounded up to $0.0001, and SVG is a flat $0.035 per image. Grok Imagine Image 2.0 charges $0.03, $0.04, $0.045 or $0.06 per image by resolution and quality, with no charge for reference images.

Resolution and sizes — GPT Image 2.6 takes aspect_ratio with 1K, 2K or 4K, a preset, auto, or an exact WIDTHxHEIGHT (multiples of 16, longest edge up to 3840, ratio up to 3:1, 0.66 to 8.29 megapixels). A 16:9 request delivered 1360×768 at 1K and 2720×1536 at 2K. Grok Imagine Image 2.0 offers 1K and 2K across fourteen aspect ratios, including 19.5:9, 9:19.5, 20:9 and 9:20. A 19.5:9 request returned 1248×576.

Quality — GPT Image 2.6 accepts low, medium, high or auto and defaults to medium on apimodels.app. The model's own default is high, which uses more tokens. Grok Imagine Image 2.0 accepts low or medium and also defaults to medium, so leaving the field out buys the medium tier.

Inputs — GPT Image 2.6 takes up to 16 reference images. It also accepts either a markup image or a mask (white marks the area to change), but not both in one request. Grok Imagine Image 2.0 takes up to three references via image_url, image_base64 or an image_urls array, addressed as <IMAGE_0>, <IMAGE_1> in the prompt. A fourth reference is rejected with an explicit error.

Outputs — GPT Image 2.6 returns png, jpeg or webp, transparent backgrounds, animated GIF and SVG, with up to 10 images per call. Grok Imagine Image 2.0 returns PNG without an alpha channel. GIF, SVG and the number of images per call are not documented for it.

Speed — In our 5 November 2026 tests, GPT Image 2.6 took about 20 seconds at 1K low and about 27 at high quality. A 1K low edit with one reference took 22 to 58 seconds, and 2K and 4K images took 13 to 25 seconds. No measured timing for Grok Imagine Image 2.0 is documented.

Limits — GPT Image 2.6 prompts run up to 32,000 characters, SVG is text-to-image only with one image per request, and results are kept for 30 days. Grok Imagine Image 2.0 tops out at 2K and three references. Its result retention is not documented.

Per-token billing against four fixed tiers

This is the difference you feel in a budget. With Grok Imagine Image 2.0, the price is known before you send the request. It is one of four numbers, set by 1K or 2K and low or medium, and references are free.

GPT Image 2.6 charges for what each request consumes. A longer prompt adds text input tokens, and each reference image adds image input at $8 per 1M. High quality, 4K and GIF use more tokens. Our measured requests (16:9, one image) show the spread:

1. 1K low: about $0.0077
2. 1K high: about $0.0162
3. 1K low edit with one reference: about $0.0157
4. 2K low: $0.0283
5. 2K medium: $0.0516

An upper bound is held when the request is accepted. The unused part is released when the image is ready.

Completed png, jpeg and webp requests return a usage object. Its charged field is the amount actually deducted, so you can log the real cost of each image. GIF, markup or mask edits and SVG return no usage object. If whoever signs off the budget wants one number per image in advance, Grok gives it. If you can work from measured costs that vary, GPT Image 2.6 came in below the matching Grok tier in each case we measured.

Output types only one side has

GPT Image 2.6 produces formats that are not documented for Grok Imagine Image 2.0. Set output_format to gif for an animation of 2 to 64 frames at 20 to 2000 ms each. All frames share one sheet of at most 3840 px on the long edge. With aspect_ratio plus 1K, 2K or 4K, apimodels.app shrinks the frames to fit.

Set output_format to svg for a static vector. SVG is text-to-image only, one image per request, at a flat $0.035 whatever the quality. The file has no scripts, external resources or embedded raster, so it suits icons, logos and flat illustrations rather than photographic scenes.

Transparency is the sharpest split. GPT Image 2.6 includes transparent backgrounds. Grok Imagine Image 2.0 cannot output an alpha channel. Asked for one, it paints the grey-and-white checkerboard as ordinary pixels in an RGB PNG, which looks like a cutout in a thumbnail but is not one. For product images you composite onto other backgrounds, send those requests to GPT Image 2.6 or add a dedicated background-removal step after Grok.

Editing: redraw the scene or point at the change

Grok Imagine Image 2.0 keeps the subject and the scene across an edit but redraws the whole frame. In one test, a red teapot was recoloured to deep navy. The teapot's shape, the wood grain, the window light, the plant on the sill and the chair behind it all carried over. What moved every time was the framing: camera angle, crop and subject size. That suits recolouring, restyling, adding or removing elements and composites. It makes Grok the wrong tool when the brief says to change one thing and leave every other pixel alone.

GPT Image 2.6 lets you point at the change with a markup image drawn over the first reference, or with a mask. We have not measured how strictly 2.6 respects masks. On the previous generation, masks acted as hints rather than hard boundaries, so name the region in the prompt as well. OpenAI positions Sunburst as the tier whose edits stay scoped to the instruction across many rounds of revision.

Reference count matters for composites. Grok fuses up to three references, and in a test with three distinct objects, each stayed individually identifiable. GPT Image 2.6 accepts up to 16, each billed as image input. Our 1K low edit with one reference measured $0.0157, against $0.0077 for a 1K low text-to-image. On Grok, references are free.

Text in the image

Grok Imagine Image 2.0 has a measured typography result. On a 2K poster with an exact headline, a date line and a small-print URL, all three came back legible and character-for-character correct. No equivalent typography test for GPT Image 2.6 is documented. Sunburst is positioned as the tier that spends more time per image on detail such as fine type, fabric weave and reflective surfaces.

What a job costs

The GPT Image 2.6 figures below multiply our measured requests from 5 November 2026 (16:9, one image). They are examples, not a fixed price: a longer prompt, more references or a different size changes the token count. The Grok Imagine Image 2.0 figures are its flat tier prices.

1. 100 product drafts at 1K low: Grok Imagine Image 2.0 costs $3. GPT Image 2.6 came to about $0.77 at the measured $0.0077 per request.
2. 100 single-reference edits at 1K low: GPT Image 2.6 costs about $1.57 at the measured $0.0157. Grok Imagine Image 2.0 costs $3, and still $3 with three references per edit.
3. 100 images at the highest 1K quality each offers: GPT Image 2.6 at high costs about $1.62. Grok Imagine Image 2.0 at medium, its highest tier, costs $4.50.
4. 10 hero variants at 2K medium: GPT Image 2.6 costs about $0.516. Grok Imagine Image 2.0 costs $0.60.

A set of 10 SVG icons costs $0.35 on GPT Image 2.6. Grok Imagine Image 2.0 has no documented SVG output. We have no measured GIF price to quote, so run a few before budgeting a batch. Any call can fail with a moderation refusal or an upstream error. On GPT Image 2.6, failed requests are never charged.

Which one to use

Choose GPT Image 2.6 Flare for any of these:

1. Animated GIF or SVG icons
2. Transparent backgrounds
3. 4K or exact pixel sizes
4. More than three references
5. Up to 10 outputs per call
6. Markup-guided edits

Move to Sunburst for images someone will zoom into. The token prices stay the same.

Choose Grok Imagine Image 2.0 when a fixed per-image price matters more than the lowest measured one. It also fits a 2K poster with exact text, a composite of up to three subjects, a recolour or restyle, or an ultra-wide 19.5:9 or 20:9 banner.

Neither is a safe pick for a pixel-locked local edit you have not tested. Grok re-renders the frame by design, and mask strictness on GPT Image 2.6 is not yet measured, so try a sample first. If you run GPT Image 2.5 for its fixed per-image prices, stay there unless you need what 2.6 adds. GPT Image 2.5 remains available on apimodels.app.

Calling both with one key

Both models sit behind the same endpoint and the same key on apimodels.app. To switch, change the model string and send the fields that model accepts.

# GPT Image 2.6 Flare: everyday text-to-image
curl https://api.apimodels.app/v1/images/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-image-2.6-flare", "prompt": "A ceramic mug on a linen table, soft window light", "aspect_ratio": "16:9", "quality": "low"}'
# GPT Image 2.6 Sunburst: animated GIF
curl https://api.apimodels.app/v1/images/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-image-2.6-sunburst", "prompt": "A paper boat rocking on gentle waves", "output_format": "gif", "frame_count": 9, "frame_duration_ms": 200}'
# Grok Imagine Image 2.0: two references fused in one call
curl https://api.apimodels.app/v1/images/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "grok-imagine-image-2", "prompt": "Put the lamp from <IMAGE_0> on the desk from <IMAGE_1>", "image_urls": ["https://example.com/lamp.png", "https://example.com/desk.png"], "aspect_ratio": "16:9", "quality": "medium"}'
# Each call returns a taskId; poll it with
# GET https://api.apimodels.app/v1/images/generations?task_id=YOUR_TASK_ID

FAQ