LLMSwaps

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

A meter with no default against a named starting tier

Both entries publish the model behind every price, and they answer the default question in opposite ways. On one, models are priced individually and nothing runs unless a call names it, so no default exists. On the other, a named model at a named resolution is what an untouched project gets. As of 2026-09-22.

Where the starting engine is decided on each entry. Recorded 2026-09-22.
On this pointfalSceneMixer
Who decides the engineThe caller, in every requestThe product, unless a tier is chosen
Is there a default at allNo, the concept does not applyYes, and it is named publicly
Where the effective choice persistsIn the caller's own codeIn whatever the product ships with
Can it move without youNo, your string does not changeNothing states that it cannot
Is the model named beside the priceYes, per modelYes, per tier

Inclusion rule. Both cells on a row come from the entry's own priced pages. A default is recorded when the page states one, or when a single-model menu makes the answer follow. Order. Who decides first, then where the choice persists.

1Both publish the engine, which is the rarer property

Most of this register sells tiers with unnamed models. These two both print the model beside the price, which is what lets a maker's announcement be matched to a bill. They differ only on who chooses.

That makes the pair useful for separating two virtues usually confused. Naming the engine is disclosure; deciding who picks it is product design, and either can exist without the other.

Naming the engine and choosing it are separate virtuesMost of this register sells tiers with unnamed models. Both these entries print the model beside the price, which lets a maker's announcement be matched to a bill. They differ only on who picks.The caller picksThe product picksDefault existsNo, every call names oneYes, and it is named publiclyEffective pin livesIn your own codeIn whatever shipsMoves without youNoNothing says it cannotEngine named by pricePer modelPer tier, with a versionA hard-coded identifier fails on the maker's date, and nobody edits it for you
Fig. 1 Either virtue can exist without the other, which is what makes this pair useful.

2A caller's string is the strongest pin available

Where every request names a model, nothing changes underneath a production unless somebody edits code. That is stronger than any commitment a page could make, and it transfers the whole obligation to the customer.

The failure mode is a hard-coded identifier that stops being served on the maker's date. Nobody will edit it, which the reading for that entry records.

3A named default is exposure you can at least see

An untouched project produces most delivered work, so a named starting tier concentrates risk in a place a reader can watch. An unnamed one concentrates it just as much and invisibly.

Neither entry promises the arrangement continues. The notice reading sets out why detection and warning are not substitutes.

  • 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
  • Default tier
    Wan 3.0 at 480Pwhat a new project getsSceneMixer, price list / recorded 2026-09-12

4Sources

Both readings come from the pages the entries publish themselves: fal and SceneMixer, read 2026-09-22. The column being compared is Default tier. Other pairs: fal and Pika, Runway and SceneMixer.