List models

GET /models

Which models your key can reach, and what each one is for.

curl https://api.jeantechnologies.com/v1/models \
  -H "Authorization: Bearer $JEAN_API_KEY"

{
  "models": [
    {
      "id": "jean-rec-1",
      "task": "recommendation",
      "status": "beta",
      "context_length": 512,
      "requires_tokenizer": true,
      "endpoints": ["generate", "predict", "embed"]
    }
  ]
}

requires_tokenizer tells you whether the model needs a fitted tokenizer before it can return catalog items. jean-rec-1 needs one for /generate but not for /predict or /embed, which operate on the backbone directly.

Model card

GET /models/{model_id}

curl https://api.jeantechnologies.com/v1/models/jean-rec-1 \
  -H "Authorization: Bearer $JEAN_API_KEY"

{
  "id": "jean-rec-1",
  "task": "recommendation",
  "status": "beta",
  "version": "1.0.0",
  "context_length": 512,
  "embedding_dim": 1024,
  "codebook": { "levels": 4, "resolution": "multi", "base_size": 2048 },
  "adapted": true,
  "adapted_at": "2026-08-14T09:12:03Z",
  "pretrain_domains": ["retail", "media", "marketplace"]
}

adapted is the field worth watching. false means you are calling the pretrained backbone with no exposure to your data, which is a fair baseline but not the ceiling. It flips to true once supervised fine-tuning on your sequences has landed. See Foundation models for the stages.

note

Model IDs are stable. A retrain that changes behavior ships as a new version, and a change that would break your integration ships as a new id.

Model card​

Model card