Four things are called Grok Imagine. Which one do you actually want?

"Grok Imagine" is a name, not a model. On LUVI it covers three separate image models and a video line, twelve models in total, and the differences are the kind that cost you a render if you guess. On September 22, 2026 a single image quoted 40 credits on the base image model, 100 on the quality tier and 120 on Image 2.0, while an eight-second video came to 1,120 credits at 720p. Here is what each one is for.

Grok Imagine Image: the cheap draft

The base image model is the one to reach for when you need to see an idea quickly. At 40 credits an image it is among the cheapest in the catalog, it generates up to four images in one request, and it offers thirteen aspect ratios, including phone shapes like 9:19.5 and 20:9.

Its dialect is the older Grok image dialect, and two rules define it: prompts run about 30 to 80 words, and supplied media is addressed with angle-bracket tokens rather than in prose. The edit model on the same line costs 44 credits.

Grok Imagine Image Quality: the same dialect, more care

The quality tier speaks the same dialect and takes the same parameters—same thirteen ratios, same one-to-four images—at 100 credits instead of 40, with the edit at 120.

There is one thing it does not do, and it surprises people: like the base model, it renders at 1k only. If you need more pixels from this family, the quality tier is not where you find them.

Grok Imagine Image 2.0: the typographer

Image 2.0 is a different family, not a newer version of the two above, and it is the only one of the three that reaches 2k. It is built for work with words in it—posters, packaging, app screens—and it plans typographic hierarchy from the brief.

Two of its rules invert against the older rows, which is exactly why "newer version" is the wrong mental model:

• References are positional, in prose—"the product from image 1 onto the surface in image 2"—not angle-bracket tokens.
• There is no word-count target. The 30-to-80-word band belongs to the older models; here the length follows the specification.

It also has a quality control the others lack: low at 80 credits against medium at 120, with the low tier roughly eight times faster. Adding 2k costs 40 credits on top.

Grok Imagine Video: the one that cannot be silent

The video line runs as text-to-video, image-to-video, reference-to-video, editing and extension, plus a v1.5 image-to-video model. Its defining trait is that picture and sound co-generate with no audio parameter anywhere in the contract: a prompt that does not name its sound gets whatever the model invents.

Clips run 1 to 15 seconds at 480p or 720p—800 and 1,120 credits at eight seconds—and reference-to-video takes up to seven images with durations capped at ten seconds rather than fifteen.

Choosing in one line each

• Trying an idea, or generating four variants of it: the base image model, at 40 credits.
• The same idea, rendered with more care, still at 1k: the quality tier, at 100.
• Anything with words in it, or anything that needs 2k: Image 2.0.
• Anything that moves: the video line—and write the audio clause, because there is no switch.

The habit that causes the most waste is carrying one row's prompt into another. The feather button beside the prompt box applies the guide of whichever model is selected, which is what converts an angle-bracket token into prose when you move from the older image rows to 2.0.

How we checked

• Models: Grok Imagine Text-to-Image, Grok Imagine Quality Text-to-Image, Grok Imagine Image 2.0 Text-to-Image and Grok Imagine Text-to-Video, with the edit, reference, extension and v1.5 models read from the catalog.
• Settings: one image at 16:9 on each image model; an eight-second video at 480p and 720p.
• Date and what was read: September 22, 2026—live credit quotes and the models' parameter contracts in LUVI, LUVI's own prompting guides for the three families, and xAI's Imagine API documentation.
• Results: 40, 100 and 120 credits for one image on the base, quality and 2.0 models; 80 credits for 2.0 at its low tier and 160 at 2k; 800 and 1,120 credits for eight seconds of video at 480p and 720p. Only Image 2.0 offered a resolution above 1k, and no video model offered one above 720p. No outputs were generated for this post, so it makes no claim about output quality.
• Drawbacks: the dialect differences come from LUVI's prompting guides rather than published vendor ablations, and we did not compare the base and quality tiers on the same prompt to see what the extra 60 credits buys.

In LUVI

Open the Grok Imagine family page and check which of the twelve models you have actually selected before you write—the names are close enough that the wrong one is easy to leave in place. Then let the feather button rewrite the prompt for that model, and set the aspect ratio in the controls rather than the prose.

New accounts start with 200 credits, and verifying a phone and a card adds up to 1,500 more. You can create a LUVI account and work out which row suits your kind of picture.

Sources

• Imagine API overview — xAI, official developer documentation. Accessed September 22, 2026.