Skip to content

DCLM-baseline

A 4T-token benchmark dataset for language models

Visit official site
Category AI models

What DCLM-baseline is

The dataset is a pretraining corpus for language model benchmarks curated from Common Crawl.

What you can do with it

  • Study curation effects on performance
  • Run model performance tests

Getting started

  • Open the Hugging Face dataset page
  • Read the guide, download it and prepare compute

Pros & cons

✓ What we like

  • A benchmark for model performance
  • Shows how curation affects training

! What to watch out for

  • Stated for research use only
  • Not for production or domain-specific training

FAQ

How large?

The material states 4T tokens and 3B documents.

For production?

Research use only, not production or domain-specific training.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review