Hub Stats: Metadata for 1M+ Hugging Face Models and ArXiv Papers

Hub Stats provides metadata for 1 to 10 million Hugging Face entities, including models and research papers, curated by cfahlgren1 with updates as recent as March 2026. The dataset aggregates information from the Hugging Face API, featuring specific splits for ArXiv papers and daily research feeds.

Use Cases

• Analyzing model fine-tuning trends by querying the baseModels field
• Cross-referencing research papers with model weights using the arxiv_papers split
• Monitoring daily research output via the daily_papers.parquet feed

Strengths

• Contains 1M to 10M records
• Includes explicit base model relationship mapping
• Sourced directly from Hugging Face API endpoints

Limitations

• Relies on user-generated tags for model lineage which may be incomplete
• Temporal snapshot that may miss models released after the last update

Provenance

Source

Hugging Face API

Collection Method

Aggregated via API

Time Range

Up to March 2026

Freshness

Updated as of March 2026.

Geography

Global