Divide by 1.25: find the provider list price behind a Sume price

Sume bills the provider list price multiplied by 1.25, so to recover the list price behind any Sume total you divide by 1.25. An 8-second Gemini Omni Flash 1.1 clip at 1080p costs $1.50 on Sume, which means a list of $1.20 and a Sume margin of $0.30.

The docs state the rule per model: at submit, Sume reserves provider list x 1.25, and the usage cost on a completed job is the Sume billable amount. The same factor applies to Video Router models, including MiniMax H3 Max.

Worked table

Each customer price comes from the price tool for a 10-second job divided by 10 (a single image for Nano Banana). List = customer price / 1.25. Margin = customer price minus list. For per-second models, multiply every column by the seconds to get a clip.

Check it against a published list

The Image API docs publish a vendor list for one model: Ideogram 4.5 at $0.03, $0.06 or $0.22 per image by quality. Multiplying by 1.25 gives the Sume amounts $0.0375, $0.075 and $0.275. The price tool agrees at medium quality: a medium Ideogram 4.5 image costs $0.075 on Sume, which is $0.06 x 1.25.

Use it in a budget

For a planning spreadsheet, store the Sume rate, not the list: it is what leaves your wallet. Use the list only to check that a quoted vendor price and the Sume price relate by 1.25, and to see how a vendor price change would flow through. A $0.10 list move per second changes the Sume rate by $0.125 per second.

Token-priced models (Seedance) follow the same rule, but the list depends on the request's resolution, seconds and aspect ratio, so estimate the exact request rather than extrapolating from one row.

LIST_TO_SUME = 1.25

def sume_price(list_usd: float) -> float:
    return round(list_usd * LIST_TO_SUME, 6)

def implied_list(sume_usd: float) -> float:
    return round(sume_usd / LIST_TO_SUME, 6)

print(implied_list(1.25))   # 10 s of Wan 3.0 at 720p -> 1.0

Gotchas

• Rounding differs: the tool rounds some token-priced totals to three decimals, so divide the displayed total, not the per-second rate you derived from it.
• The catalog's pricing_skus on GET /v1/videos/models is the machine-readable price, so prefer it to a hand-copied table.
• Image quality changes the list for some models: Ideogram 4.5 steps from $0.03 to $0.22 per image, GPT Image 2.5 uses token rates, so quote the quality with every price.