LLMSwaps

Which model each app runs today, and when it last changed

A bundled subscription against models priced one by one

Both entries sell other companies' models and neither states a default. On one they are bundled into a subscription, so which engine runs is unstated and unpublished. On the other each model carries its own price and nothing runs unless a request names it. As of 2026-09-22.

How each arrangement settles which engine produces the work. Recorded 2026-09-22.
On this pointinvideo AIfal
What a customer buysA subscription covering six makersOutput from a named model
Who chooses the engineUnstated on the pageThe caller, every time
Is a version givenMakers named, versions mostly notModel names on the price list
Can the engine be identified laterNot from anything publishedYes, from the invoice
What a version change reachesUnknowable from the pageOnly callers naming that model

Inclusion rule. Both cells on a row come from the entry's own priced pages. A maker counted here is one the entry itself names. Order. What a customer buys first, then who chooses, then what is recorded.

1A bundle hides the answer; a line item is the answer

Where six makers sit behind one subscription, nothing published says which produced a particular clip. Where each model is priced separately, the bill records it without anybody making an effort.

That is the single most useful difference in this column. Every explainer here ends by recommending a per-file record, and one of these two arrangements produces it automatically.

A bundle hides the answer; a line item is the answerWhere six makers sit behind one subscription, nothing published says which produced a particular clip. Where each model is priced separately, the bill records it without anybody making an effort.A requestNames a model, orreaches whatever thesubscriptionprovides.ThenThe billOne line per model,or one subscriptioncovering six makers.LaterThe questionWhich engine madethis clip?Answerable, or not.Both inherit other companies' schedulesOnly the priced list lets a dated announcement be matched to a purchased line.
Fig. 1 Every explainer here recommends a per-file record, and one of these arrangements produces it automatically.

2Both inherit other companies' schedules

Neither entry makes the models it sells, so both are exposed to retirements decided elsewhere. One carries six makers' schedules with no versions; the other carries several with names on a price list.

Only the second lets a dated announcement be matched to a purchased line. The aggregation question sets out why that matching is the whole problem.

3Neither has a default to publish

One has none because a subscription does not expose the choice; the other has none because the choice is compulsory. The first is an absence of disclosure and the second an absence of the concept.

The register records them separately for that reason. The reading for the priced list explains why a caller's own string becomes the effective default.

  • Models carried
    Google Veo 3.1, Sora 2, Kling AI, Wan AI, Pixverse AI and Hailuo AIaggregatedinvideo AI, model list / recorded 2026-09-12
  • Models carried
    Wan 2.5, Kling 2.5 Turbo Pro, Veo 3 and Ovi, each priced individuallypriced per model rather than per planfal, pricing page / recorded 2026-09-22

4Sources

Both readings come from the pages the entries publish themselves: invideo AI and fal, read 2026-09-22. The column being compared is Default tier. Other pairs: Wan-AI and fal, Foundry and Bedrock.