SEC EDGAR: Financial Filings Corpus with 43 Billion Tokens

A corpus of 8 million samples and 43 billion tokens extracted from all major filings in the U.S. Securities and Exchange Commission's EDGAR database. The dataset was created by Datamule, Teraflop AI, and Eventual using the datamule-python library and API. It comprises 590 GB of data.

Use Cases

• Train language models for financial document understanding based on the large-scale text corpus.
• Analyze corporate disclosure trends and patterns across millions of filings.
• Develop information retrieval systems for financial regulatory documents.
• Perform named entity recognition and relationship extraction on corporate filings.

Strengths

• 590 GB of data provides substantial volume for model training.
• 8 million samples and 43 billion tokens indicate a large-scale text corpus.
• Covers all major filing types from the authoritative SEC EDGAR database.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2026-04-15 09:12:01; freshness should be verified.

Provenance

Source

U.S. Securities and Exchange Commission (SEC) EDGAR database.

Collection Method

Collected using the datamule-python library and official Datamule API.

Freshness

Last updated 2026-04-15 09:12:01.

Geography

United States (primary).