Wan 3.0 leads text-to-video at 1,156 Elo: $12 listed, $15 on Sume

Wan 3.0 is first on Artificial Analysis's text-to-video board (AA-Video-T2V v2.0) with 1,156 Elo, and the board lists it at $12.00 per minute. On Sume you call it as wan-3.0. At 1080p it bills $0.25 per second, which is $15.00 per minute, because Sume bills the provider list price times 1.25 on every video model.

That is the whole answer for a buyer: the leaderboard price is the list price, and the Sume price is list plus the 25% house margin. The rest of this page shows the arithmetic at all three resolutions and the limits that decide whether Wan 3.0 fits your clip.

What the leaderboard says

The board ranks models by Elo from blind pairwise votes of a human panel, on text-to-video with audio. The first four rows on the day this page was written:

What Wan 3.0 costs on Sume

Alibaba's Model Studio page lists Wan 3.0 at $0.05, $0.10 and $0.20 per second for 480P, 720P and 1080P in Singapore. Sume's catalog uses the same per-second list and multiplies it by 1.25. The $0.20 per second 1080p list is $12.00 per minute, which matches the leaderboard figure.

Limits that matter more than a point of Elo

Sume's video docs give wan-3.0 a duration range of 2 to 30 seconds. It takes text, first and last frame images, and reference images, videos and audio. A leaderboard score says nothing about these limits, so check them against your clip before you read anything into the rank.

• Length: 2 to 30 seconds in one job, per Sume's video docs. Alibaba's page also lists a 30 second maximum.
• Resolution: 480p, 720p and 1080p. There is no 4K on this id.
• Price rule: Sume reserves list x 1.25 when you submit, so the wallet must cover the whole clip first. A 30 second 1080p job reserves $7.50.

When to pick it

Pick Wan 3.0 when the clip is long (15 to 30 seconds), when you want the best blind-vote score for a prompt-only clip, or when the per-minute price matters. A 720p draft costs $0.625 for 5 seconds, so you can try three prompts for less than $2 before you spend $2.50 on a 10 second 1080p final.

Pick something else when you need 4K (Gemini Omni Flash 1.1 lists it, at 3 to 10 seconds). The Elo gap between first and third is 13 points, so the price and the limits are the better tie-breakers. See the Seedance 2.5 comparison for that case.

Budget a batch before you submit

Sume reserves the provider list x 1.25 for the whole job when you submit, and returns 402 insufficient_credits if the balance is short, so cost a batch first. Twenty 10 second clips at 720p are 20 x 10 x $0.125 = $25.00, and the same batch at 1080p is 20 x 10 x $0.25 = $50.00. At 480p it is 20 x 10 x $0.0625 = $12.50.

Run a small slice first. Five clips at 720p cost $6.25, enough to see whether the board's first place carries over to your prompts before you reserve the rest.

• Resolution changes the price by a flat multiple: 720p is twice 480p, and 1080p is twice 720p.
• The clip length range is 2 to 30 seconds, so a single job can cover most ad cuts.