Data Sources

AI Coding Stack combines official vendor documentation with selected open data sources. These sources inform the catalog, but every record remains subject to project review.

Source priority

  1. Official product documentation, model cards, release notes, and pricing pages
  2. Publisher-maintained model repositories and model cards on Hugging Face for model artifacts, licenses, base-model relationships, supported libraries, datasets, and evaluation metadata
  3. Models.dev for catalog discovery and cross-checking model specifications, capabilities, context limits, release dates, and provider pricing
  4. Artificial Analysis for the Intelligence Index, benchmark methodology, and comparable API market observations

When sources disagree, an applicable first-party source takes precedence.

Hugging Face

Hugging Face Hub model repositories and model cards provide information about open-weight model artifacts, licenses, base-model lineage, supported libraries, tasks, datasets, and evaluation results. AI Coding Stack prefers repositories maintained by the model publisher. Community uploads are treated as supplementary evidence and do not by themselves establish vendor claims or current commercial pricing.

Models.dev

Models.dev is an open-source database of AI model specifications, pricing, and capabilities. AI Coding Stack uses it as a reference when reviewing model and provider records. Values are checked before they are incorporated and may be supplemented or overridden by official sources.

Artificial Analysis

The model intelligence rankings use the Artificial Analysis Intelligence Index and link to its methodology. The price × intelligence view primarily uses official API pricing recorded in the catalog; where a comparable first-party USD price is unavailable, it may use a comparable API offer tracked by Artificial Analysis, as disclosed on that page.

Verification and freshness

Verified records include their cited sources and latest review date whenever available. Specifications, prices, and benchmark results can change, so they should be read together with those source links and dates.

Corrections and source improvements are welcome through the AI Coding Stack repository.