Grok Imagine API pricing: xAI 8c and 2c a second, Sume 1.25c

Grok Imagine Video 1.5 costs $0.08 a second on xAI's own price table and 1.5 Lite costs $0.02 a second. On Sume the one Grok row, grok-imagine-video-1.5, bills 1.25 cents a second at both 480p and 720p, so a 10-second clip is about 13 cents. Sume has no Lite row, so the 2-cent comparison is a xAI-only price.

The two price lists

The first table is xAI's. The xAI models page gives one per-second figure for each model, and the video guide says that duration and resolution both affect the total cost. This post only quotes the headline figures on that page, not a per-resolution ladder. If you need the exact figure for a resolution, read the xAI pricing page for your account before you budget.

The second table is Sume's. Sume bills the provider list price times 1.25, and the repo's Grok estimator records a provider list of $0.01 per output second, which gives $0.0125 a second. The billable amount is ceiled to the cent for each job, so short clips round up by a fraction of a cent.

What the Sume row is, and what it is not

The two lists are not the same product. Sume's row is the image-to-video route of Grok Imagine 1.5 through its provider, and it takes one start image. It does not take a text-only prompt, an end frame, reference media or an aspect_ratio. It accepts 4 to 15 seconds at 480p or 720p. xAI's own guide describes more modes than that, including video editing, extension and reference-to-video, and those are not exposed on Sume's Grok row.

Check the live numbers for your key with GET /v1/videos/models. The catalog's pricing_skus values are the Sume billable rates, so you can read them rather than trust a blog table.

Which one to call

Use the xAI figures when you call xAI directly and need Lite or text-to-video. Use the Sume row when you already have a still and want one key, one wallet and the same job API as the other video models. For a text prompt with sound, pick a different row, for example Gemini Omni Flash 1.1, which Sume lists at 3 to 10 seconds with native audio.