Parameter Count (ML Models)

Ask about Parameter count

Ask your first 2 questions free below. Subscribers get 150 a day for $29 a month.

Ask CASRAI answers research-administration questions and cites the passages behind every claim. When our sources don't cover a question, it says so.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays free to read, with no account and no card.

Parameter count is the total number of learnable scalar weights in a machine-learning model, conventionally reported as a shorthand count such as “7B” (7 x 10^9 parameters). It is one of the most basic disclosed metadata fields for a model, alongside architecture family and training-data description, and is frequently used — imperfectly — as a rough proxy for a model’s capacity or expected capability.

Why it is a weaker proxy than it looks

Parameter count alone does not determine capability: a mixture-of-experts architecture may have a large total parameter count while only activating a small fraction of those parameters per query, making raw parameter count a poor comparator against a dense model of similar “active” size. Training-data quality and quantity, and increasingly inference-time compute for reasoning-style architectures (see inference), can matter as much as, or more than, parameter count for a given task. Reporting parameter count alongside training compute (FLOPs) and architecture type gives a materially more informative picture than parameter count alone.

Why it matters for research and procurement

Parameter count is one of the few model-scale figures that vendors of closed, API-served models routinely decline to disclose, which is itself worth noting in a methods section or procurement record: “parameter count not disclosed by vendor” is a real limitation to record, not an omission to paper over.

References

See also: Mixture-of-experts (MoE), Compute (FLOPs estimate), Training data composition.

Use in your systems

JATS XML <role> element

xml

<role vocab="credit"
      vocab-identifier="https://casrai.org/dictionary/"
      vocab-term="Parameter count"
      vocab-term-identifier="https://casrai.org/dictionary/term/parameter-count" />

Schema.org DefinedTerm (JSON-LD)

json

{
  "@context": "https://schema.org",
  "@type": "DefinedTerm",
  "@id": "https://casrai.org/dictionary/term/parameter-count",
  "name": "Parameter count",
  "identifier": "https://casrai.org/dictionary/term/parameter-count",
  "description": "The total number of learnable scalar weights in a machine-learning model, conventionally reported as a count (e.g., 7B = 7 x 10^9 parameters) and disclosed as a basic model metadata field.",
  "inDefinedTermSet": "https://casrai.org/dictionary/domain/ai-ml-research-outputs#set",
  "url": "https://casrai.org/dictionary/term/parameter-count",
  "alternateName": [
    "model size (parameter count sense)"
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "publisher": {
    "@id": "https://casrai.org/#organization"
  },
  "author": {
    "@id": "https://casrai.org/#editorial-team"
  },
  "datePublished": "2026-05-21T02:22:51",
  "dateModified": "2026-09-05T14:24:42",
  "inLanguage": "en-GB",
  "isAccessibleForFree": true
}

Parameter count