Skip to content
Get API key

Models and versioning

Mouse serves one model family (mouse-1) behind immutable, date-stamped snapshot IDs. Every decision response echoes the exact snapshot ID that scored the request in both the JSON body (model) and the Mouse-Model-Version HTTP header.

Query GET /v1/models to inspect all active snapshots and alias mappings:

AliasResolves to snapshotMax state tokensModalitiesStatus
mouse-1mouse-preview-2026-10-1316,384["text", "image"]Preview
mousemouse-preview-2026-10-1316,384["text", "image"]Preview

You can pass either an alias ("mouse-1", "mouse") or the immutable snapshot ID ("mouse-preview-2026-10-13") in the model field of POST /v1/decisions and POST /v1/calibrations.

  1. Immutable snapshots: Once a snapshot ID such as mouse-preview-2026-10-13 is published, its weights, tokenizer, rendering format, and scoring numerics never change.
  2. Always echoed: Even when you request model: "mouse-1", the response body's model field and the Mouse-Model-Version header always return the concrete snapshot ID ("mouse-preview-2026-10-13"). Log this value alongside your application telemetry.
  3. When to pin a snapshot ID: If you fit a calibration profile (calp1.) or tune custom confidence thresholds against your validation set, pin model: "mouse-preview-2026-10-13" in production. Calibration profiles are bound to the snapshot they were fitted on and return 409 profile_model_mismatch if used with a different snapshot.
  4. 30-day alias notice & 12-month support window: Aliases (mouse-1, mouse) advance to newer snapshots with at least 30 days' advance notice. Superseded snapshots remain live for at least 12 months after their successor ships, with retire_after populated in GET /v1/models.

A tuning produces a model of your own, named after the suffix you chose:

FormExampleBehaviour
Tuned idmouse-1:ft-acme-7f3a2cImmutable. The six hex characters identify the job; the weights never change.
Tuned aliasmouse-1:ft-acmeThe newest ready tuning on your account with that suffix. Moves when a later job with the same suffix publishes (and, with baseline set, only if it beat the previous one).

Both are accepted in model on POST /v1/decisions and POST /v1/calibrations, and both are listed in GET /v1/models for your account only. Whichever you send, the response's model field and the Mouse-Model-Version header echo the full tuned id, so a log line always names the exact weights that answered. Pin the tuned id in production when a calibration profile is involved; profiles bind to tuned ids exactly as they bind to snapshots. A deleted tuning returns 404 model_not_found, and so does an alias with no ready tuning left behind it.

Tuned models follow the base snapshot they were trained on. When the base model moves to a newer snapshot, existing tunings keep serving on the snapshot they were built against for the same 12-month support window; post your examples again to tune the new base.

mouse-1 is a proprietary decision model purpose-built for agent-harness control: tool routing, action-gate safety, option-order invariance (<1% order flip rate), and resistance to planted prompt injections in untrusted state (0% flip rate on held-out injection suites).