Is Grok Imagine cheaper than GPT Image 2.5? Only versus high

No, not at the tiers most teams use. On Sume, Grok Imagine costs a flat $0.0250 per image, while GPT Image 2.5 at 1024x1024 costs $0.0074 at low and $0.0165 at medium, both below Grok. Grok Imagine becomes the cheaper of the two only against high at $0.0659, xhigh at $0.1171 and max at $0.2635.

The question is really about which GPT Image 2.5 tier you would otherwise run, and about how much of the Grok price is fixed, because Grok's price does not move with size or quality while GPT's moves by a factor of 36 between low and max.

Per-tier comparison at 1024x1024

The third column compares the GPT price with Grok's $0.0250. All GPT figures are Sume quotes with a one-character prompt.

Where Grok Imagine's flat price helps

Grok Imagine's price does not depend on the output size, so large frames cost the same as small ones. At 3840x2160 the GPT Image 2.5 medium quote is $0.0325, still above Grok's flat figure, and at high it is $0.1251, five times as much. If you need large frames from a single call at a low flat price, Grok is the better row.

Sume's catalog lists Grok Imagine with edit support and one output image per call, against four for GPT Image 2.5. A job that needs four takes means four Grok calls, so the price per call has to be multiplied before you compare. Four Grok calls cost $0.1000, and four GPT Image 2.5 medium images in one call cost $0.0660.

What changes the answer

Three things move the break-even.

• Size. A GPT Image 2.5 medium image stays below Grok up to 2560x1440 ($0.0180), but at 3840x2160 it passes it. Leaving image_size out reserves the upper bound and quotes $0.0556 at medium.
• References. Each reference image added to a GPT Image 2.5 request adds about a quarter of the output price. At medium one reference moves the quote from $0.0165 to $0.0209, still below Grok.
• Release pace. xAI publishes changes on its release notes page (read 2026-10-05); Sume's catalog price is what you pay, so recheck GET /v1/images/models after any update.

Decision

If the alternative is low or medium GPT Image 2.5, use GPT Image 2.5 and save money. If you are comparing against high or above, Grok Imagine saves 62% at high and 79% at xhigh, but you give up the six-step quality setting.

Run the same ten prompts through both and score them before you decide. The price table says which row is cheaper, and your prompts say whether the cheaper row is good enough.

A mixed setup is also reasonable. Use Grok Imagine for a high-volume feed where a flat price makes the budget easy to predict, and keep GPT Image 2.5 at medium for the images that need references, transparency or a specific quality tier. Both are called through the same POST /v1/images request, so the split is a routing decision in your own code.

Whichever you pick, set a spending alarm from usage.cost. The response reports the amount billed for every call, so summing it per day catches a runaway loop within hours and not at the end of the month. Grok's flat price makes that sum trivial to forecast, and GPT Image 2.5's tier-driven price makes it worth watching more closely.