AA text-to-video top 12: six you can call on Sume, six you cannot

Of the 12 models on the AA text-to-video board, six have a Sume id that takes a prompt-only request. The list, with Elo and Sume's price per minute.

5 min readSume
All posts

Six of the twelve models on Artificial Analysis's text-to-video board (AA-Video-T2V v2.0) can be called from Sume with a prompt and no image: Wan 3.0, Seedance 2.5, MiniMax H3, MiniMax H3 Max, Seedance 2.0 and Gemini Omni Flash. The other six are either not in Sume's video id list or, in one case, listed for image-to-video only.

A leaderboard is a list of models that exist. A buyer needs the list of models that exist where they can send a request. This page lines the two up.

The twelve rows

Elo and listed prices are from the board. The Sume column names the id in Sume's video list. The last column is Sume's price per minute at the resolution shown, list x 1.25.

AA-Video-T2V v2.0, top 12, against Sume's video ids, read 2026-10-08
RankModelEloListed per minutePrompt-only on SumeSume per minute
1Wan 3.01156$12.00wan-3.0$7.50 at 720p
2Utopai X1149No APINot listedn/a
3Dreamina Seedance 2.51143$34.12seedance-2.5$34.67 at 720p
4MiniMax H3 (768p)1137$4.80minimax-h3$4.50 at 768p
5MiniMax H3 Max1130$4.80minimax-h3-max$6.00 at 768p
6FLUX 31128$17.40Not listedn/a
7Dreamina Seedance 2.01116$22.45seedance-2$22.68 at 720p
8Gemini Omni Flash 1.11113$9.12gemini-omni-flash-1.1$7.50 at 720p
9SkyReels V41074$21.00Not listedn/a
10Agnes-Video-2.51057$2.40Not listedn/a
11Grok Imagine Video 1.51045$15.00Image-to-video onlyn/a
12HappyHorse-1.11042$10.80Not listedn/a

What the Sume column does not say

A Sume id is not a promise that the board's score carries over. The board's prices are the vendors' list prices. Sume's prices include the 25% margin, so the same model can be a little more expensive, as with Seedance 2.0 ($22.45 listed against $22.68 on Sume). H3 is the other way round: the board lists $4.80 and Sume's 768p rate is $4.50, so the two sources use different bases. Treat the two price columns as separate sources.

Where the six differ in practice

The six ids do not share one envelope. Read capabilities from GET /v1/videos/models before you plan a batch, and use the duration range to rule models out first.

  • Wan 3.0 and Seedance 2.5: up to 30 seconds. Wan starts at 2 seconds and Seedance at 4.
  • MiniMax H3 and H3 Max: 5 to 15 seconds. 768p is native; it is not 720p.
  • Seedance 2.0: 4 to 15 seconds, with 480p, 720p and 1080p.
  • Gemini Omni Flash 1.1: 3 to 10 seconds, in 16:9 or 9:16 only, with native audio always on.

What to do with the six

Send one prompt to the three cheapest (H3, H3 Max, Wan 3.0 at 720p) and one to the leader of your quality bar, then compare the clips. A 5 second run of all four costs $0.375 + $0.50 + $0.625 + $2.889, which is $4.39. If you want a method, the Wan 3.0 board post and the bake-off recipe both apply.

Sources

Related posts

More in Models

All Models posts

Written by Sume