Models & datasets

The catalog endpoints: the base models you can grow, your trained models (and how to download them), and your datasets. For the upload walkthrough see Datasets; to retrieve a finished model see Monitor & retrieve.

Base models

GET /base-models

The shared catalog of models you can grow. Use an id here as base_model_id on a training run.

[ { "id": "42c01c10-…", "family": "falcon", "name": "falcon-h1-1.5b",
    "version": "bf16", "params": 1550000000, "precision": "bf16" } ]

Your models

GET /models

Your trained output models, newest first.

[ { "id": "…", "name": "falcon-h1-x2-chassis", "version": "v1",
    "params": 2360000000, "base_model_id": "…", "lineage": { "job_id": "…" },
    "status": "ready", "created_at": "…" } ]

GET /models/{id}/download

A download manifest — a presigned GET URL per file. The console never proxies bytes; you pull the weights directly from object storage. URLs are short-lived; re-request the manifest if they expire.

{ "model_id": "…", "name": "falcon-h1-x2-chassis", "version": "v1",
  "files": [
    { "rel": "model-00001-of-00002.safetensors", "size": 4718592000, "url": "https://…signed…" },
    { "rel": "config.json", "size": 1024, "url": "https://…signed…" }
  ] }

curl https://console.axomlabs.ai/api/models/<id>/download -H "Authorization: Bearer $AXOM_KEY"
curl -o model-00001.safetensors "<url-from-manifest>"

Datasets

GET /datasets

Your datasets, newest first.

[ { "id": "…", "name": "wikitext-x2-chassis", "version": "v1",
    "shard_count": 8, "bytes": 2400000000, "status": "ready", "created_at": "…" } ]

POST /uploads/presign

Mint a presigned PUT URL to upload a dataset file directly to storage — the console never receives the bytes. Full flow: Datasets.

// request
{ "filename": "train.jsonl", "content_type": "application/jsonl" }

// response
{ "upload_id": "…", "key": "tmp/…/train.jsonl", "url": "https://…signed…", "method": "PUT" }