Models

The reasoning model behind every SwishX request — the thing that decodes briefs, harnesses context, links claims, routes to a downstream generator, and verifies output — is what people mean when they say "the SwishX API."

SwishX names each model generation after a Nobel laureate in Physiology or Medicine, rotating to a new name every two or three versions:

Current version

finsen-3.2

Pass this as the model field on any request. Sharper claim-grounding accuracy and lower multi-model routing latency than 3.1 — see the changelog for what changed.

Choosing a version

Most integrations should just use the current version and move forward as new ones ship — the request/response shape has been stable since Finsen 3.0. Pin an older version only if:

• You've validated a specific version's compliance behavior for a submission already in flight, and don't want generation behavior to shift mid-review.
• You're comparing output quality across versions before upgrading a production integration.

curl https://api.swishx.com/v1/videos \
  -H "Authorization: Bearer $SWISHX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "finsen-3.1",
    "prompt": "..."
  }'

Keyframe model

Video generations that use keyframe-first rendering (see Model routing) also invoke a paired image model for the opening still, finsen-3.2-keyframe. You don't call this directly — it's invoked automatically as part of a /v1/videos request — but its own rate limits are tracked separately.