Seedance 2.5 Fast API: Sume has no such id; prices of the four

Searching 'seedance 2.5 fast'? Sume's registry has seedance-2.5, seedance-2, seedance-2-fast and seedance-2-mini. Prices at 480p and 720p, 5 s, 9:16.

5 min readSume
All posts

The Seedance model ids in Sume's registry on 2026-10-07 are seedance-2.5, seedance-2, seedance-2-fast and seedance-2-mini. There is no seedance-2.5-fast: a request with that id is a model-not-found error, and the repo has no price for it. If your code or a blog post says 2.5 Fast, the cheaper speed tiers on Sume are the Seedance 2.0 family.

That is a statement about Sume's catalog only. Whether another provider offers a faster 2.5 tier is not something this page claims. The catalog is the source of truth, and GET /v1/videos/models lists every id your key can use.

The four ids side by side

Prices are Sume's billable amount (provider list x 1.25), computed from the token formula ceil(width x height x seconds x 24 / 1024) at the per-1,000-token list rate. 5-second clips at 9:16.

Seedance ids and prices on Sume (read 2026-10-07)
Model idDurationsList $ per 1k tokens480p, 5 s720p, 5 s
seedance-2.54-30 s0.0214 (0.0234 at 1080p)$1.343385$2.889
seedance-24-15 s0.0140$0.87885$1.89
seedance-2-fast4-15 s0.0112$0.70308$1.512
seedance-2-mini4-15 s0.0070$0.439425$0.945

Check the live catalog in one line

This prints every Seedance id your key can submit to. If a new tier ships, it shows up here before it shows up in any article.

The 2.5 model is the only one with 30-second clips; the others stop at 15 seconds.

curl -s https://api.sume.com/v1/videos/models -H "x-api-key: $SUME_API_KEY" \
  | jq -r '.data[].id | select(startswith("seedance"))'

Picking between them

Use seedance-2-mini for the cheapest framing test, seedance-2-fast when you want a faster, cheaper pass than 2.0, and seedance-2.5 for the 30-second clips and the highest tier. At 720p and 5 seconds, 2.5 costs about 3.1x mini and 1.9x fast. Drafts on a cheap tier and finals on 2.5 is the usual pattern; one Seedance prompt is not guaranteed to look the same across tiers, so compare outputs rather than assuming.

Sources

Related posts

More in Models

All Models posts

Written by Sume