AI and ML sample files
JSON Lines, CSV and JSON datasets for testing data loaders, tokenisation pipelines and evaluation tooling.
Formats in this category
| Format | What it is | Sizes |
|---|---|---|
| JSONL sample files.jsonl application/jsonl | One JSON value per line. The usual format for logs, data pipelines and machine-learning datasets. | 1 KB to 10 MB |
| CSV sample files.csv text/csv | Tabular data as plain text, one record per line. The universal import and export format. | 1 KB to 10 MB |
| JSON sample files.json application/json | The standard data format of web APIs: objects, arrays, strings, numbers, booleans and null. | 1 KB to 10 MB |
About these files
Machine-learning tooling runs on line-oriented data: JSON Lines for fine-tuning and evaluation sets, CSV for tabular work. Before pointing a loader at real data, it is worth checking that it streams, batches and reports progress correctly on a file of known size.
These samples are structured records rather than training text. They are intended for testing the plumbing of a data pipeline, not for training a model.
Popular downloads by size
Jump straight to a file. Every size is exact to the byte.
Or browse every format at one size: 1 KB, 5 KB, 10 KB, 25 KB, 50 KB, 100 KB, 500 KB, 1 MB, 5 MB, 10 MB.
What to test
- File parser testing
Well-formed files and deliberate edge cases for testing code that reads CSV, JSON, XML, PDF and more.
- Performance testing
Large, predictable inputs for measuring throughput, memory use and processing time.
- Automated testing
Small, deterministic fixtures for Playwright, Selenium, Cypress and unit tests.
Related categories
AI and ML files: questions
- Can I train a model on these files?
- You can, but there is nothing to learn from them: the records are synthetic and random. Use them to test loaders and pipelines.