LTX 2.5 API: check the Sume catalog before you send a model id

LTX-2.5 has its own API and open weights. On Sume, send only ids that GET /v1/videos/models lists; an unknown id returns model_not_found. How to check and pick.

4 min readSume
All posts

LTX-2.5 is available from LTX's own API and as open weights, but this post only claims what the Sume docs say: send a Sume model id only if GET /v1/videos/models lists it. An unknown id returns the typed error model_not_found, so list first and pick from what comes back.

The LTX details below are from the LTX-2.5 page, read 2026-09-30. Sume's side is from the video docs and Video Router docs. This post does not state whether any LTX id is or will be listed; the catalog response is the source of truth.

What does LTX say LTX-2.5 is?

The page describes an open-weights video model with native multi-shot scenes, automatic clip duration, beta precise video editing, and 4K HDR and RAW support. API pricing is billed per second of generated video, listed by tier and resolution.

LTX API per-second prices as written on the LTX-2.5 page, read 2026-09-30: https://ltx.io/model/ltx-2-5
LTX tier720p1080p1440p4K
ltx-2-5-fast$0.09/sec$0.13/sec$0.19/sec$0.30/sec
ltx-2-5-pro$0.12/sec$0.17/secNot listedNot listed

How do I check what Sume offers?

Call GET /v1/videos/models. Each entry has id, supported_resolutions, supported_aspect_ratios, supported_durations (whole seconds), supported_input_references, generate_audio, and pricing_skus. The docs say to check these before submitting a generation request.

Do not copy LTX's tier names (ltx-2-5-fast, ltx-2-5-pro) into a Sume request unless the catalog shows that exact id. If it is absent, the request fails with model_not_found.

What should I pick if there is no LTX id?

Pick by the limits you need, not the name. Limits are per model: for example, seedance-2.5 accepts 4 to 30 seconds at 480p, 720p and 1080p. Compare supported_durations and supported_resolutions across the listed ids, then test one short clip.

If you do not care which family renders, sume/auto lets Sume choose; the family that ran is not disclosed. For a comparison of how other models differ, see one API for Kling, Seedance and other video models.

What if I need the open weights themselves?

The LTX page says LTX-2.5 can be fine-tuned and deployed on your own infrastructure. That is a self-hosting path with its own licence terms, set out on the LTX page; Sume's API is a hosted catalog and does not take your weights.

For the earlier generation, read LTX-2 and open-source video.

Sources

Related posts

More in Models

All Models posts

Written by Sume