The data

Name: YFarmX Model Identity dataset
Creator: YFarmX

Every figure on these pages renders from committed data files, and the files are public. Check our numbers, rebuild our clusters, or run the same strings against your own endpoint.

Catalogue: 422
Measured: 158
Complete fingerprints: 152
Distinct signatures: 46
Tokenizer groups: 35
Method: v1-50

The files

Dated 22 August 2026; regenerated only by the published pipeline

index.jsonThe full catalogue: 422 model records with declared metadata, measurement status, family assessment, confidence and its reason.JSON fingerprints.jsonMeasured fingerprints: 158 models' marginal token counts, signatures, cluster membership and observation dates.JSON test-strings.jsonThe 50 adversarial test strings, byte-exact and in measurement order, for reproducing any fingerprint.JSON records.csvThe catalogue flattened to one row per model for spreadsheets: maker, label, measured family, confidence, status, context, output, pricing, dates.JSON record.json, per modelEvery record page serves its own JSON at <record URL>/record.json; the link here is the Ox Alpha example.JSON

The shape of a record

{
  "id": "z-ai/glm-5.3",
  "lab": "Z.ai",
  "declared": { "tokenizer": "Other", "context": 1048576, "max_output": 131072, … },
  "measurement": { "status": "measured", "at": "2026-08-21",
                   "signature": "tk_133b086c", "counts": { "zh-long": 61, … } },
  "identity": { "kind": "lineage", "family": "Z.ai", "confidence": "very-high",
                "confidence_reason": "an exact fingerprint shared with 2 models…" },
  "neighbours": [ { "id": "stealth/ox-alpha", "same": 50, "total": 50 }, … ]
}

Abridged for shape; a full example is any record's own file, for instance glm-5.3/record.json.

Reuse

Quote the figures freely with attribution to YFarmX and a link to the record page the figure comes from. Measurements are dated and append-only, so a cited number stays checkable at its URL. For bulk or commercial reuse of the dataset, ask first.