Datasets:
Dataset Viewer
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
DragonData Finance Corpus
The world's largest open, permissively-licensed finance corpus for LLM pretraining.
Built by Dragon Limited - 100% free, publicly available sources.
Overview
| Attribute | Value |
|---|---|
| Repository | dragonlimited/DragonData-Finance-Corpus |
| Tokenizer | bigcode/starcoder2-3b (vocab 49,152) |
| Format | Binary shards (data/train-XXXXX-of-00001.bin, uint16) |
| Target | 46T tokens across 25 domains |
| Current | ~33B tokens, 246 shards |
| License | Permissive only (CC BY, CC0, BSD, Apache, MIT, Public Domain) |
| Status | Building (2026-09-01) |
Shard-Domain Mapping
| Shard Range | Domain | Source | Tokens |
|---|---|---|---|
| 00000-00001 | fine_web_finance | FineWeb (finance-filtered) | ~410M |
| 00002-00009 | mixed_domains | Various (BeanCounter, Bloomberg, arXiv) | ~20M |
| 00010-00015 | books_textbooks | Project Gutenberg | ~8M |
| 00016-00023 | mixed_domains | ESG, Central Bank, Insurance, Real Estate | ~5M |
| 00024-00033 | fine_web_finance | FineWeb (finance keyword filter) | ~60M |
| 00034-00102 | social_discourse_finance | Reddit Finance | ~35M |
| 00103-00110 | fine_web_finance | FineWeb | ~40M |
| 00111-00245 | us_court_records_full | Turkish Court Decisions | ~180M |
| 00111-00245 | social_discourse_finance | Reddit Finance | ~130M |
| 00111-00245 | economics_reference | Wikipedia | ~130M |
Note: Domain tags are embedded in each document as ##domain and ##source prefix lines.
Document Format
Each document is prepended with metadata tags:
##domain ##source
Licensing
Only permissive or public-domain sources: CC BY, CC0, BSD, Apache, MIT, Public Domain, ODC-By, OGL.
NOT included: Proprietary data, paywalled content, restrictive licenses.
- Downloads last month
- 323