Invisible models in Pika vs Sume Auto echoing sume/auto

On the new Pika, models are named when you want them and invisible when you do not. Sume has the same two modes through the API: send model: "sume/auto" and Sume picks the family, or send a bare catalog id such as seedance-2 to pin one. With Auto, the response says sume/auto and does not tell you which family ran.

Pika's wording is from its September 17 post; Sume's from Video generation, read 2026-10-01.

What does Pika say about model choice?

The post says its apps automatically use the model best suited to your task, and that if you want more control you can still choose a specific model yourself. It names Seedance, GPT Image, MiniMax H3 and Pika's own models among those it uses.

What does Sume Auto return?

The Video generation page says model: "sume/auto" lets Sume pick the family, and the poll response reports "model": "sume/auto". It states that Sume does not disclose which family served the request, and that you should not build on any observable trait of the output to infer it. Resolution is a pure function of the normalized request and the catalog version, so an idempotent replay prices and routes the same way.

When should I pin a catalog id instead?

Pin when something downstream depends on the model: a 30-second length, reference video, or a budget you computed from one model's rate. The Video Router page's own guidance is "let Sume pick" for Auto and "pin a catalog model" for a chosen one. Auto create controls default to 720p and 8 seconds, so set resolution and duration explicitly if you need otherwise; see Auto defaults.

Can I tell which model made my clip?

Not from an Auto response. If you need to know, pin a catalog id from GET /v1/videos/models. For a comparison with another router's Auto, read Runway's media router vs Sume Auto.