Skip to content

DCLM

A framework for building language model datasets

Visit official site
Category AI models

What DCLM is

The framework is described as a comprehensive setup for building and training large language models with standardised corpora and evaluations.

What you can do with it

  • Try different dataset construction strategies
  • Evaluate models with more than 50 methods

Pros & cons

✓ What we like

  • Pretraining recipes are based on open_lm
  • A leaderboard collects community model submissions

! What to watch out for

  • The workflow sets up AWS storage and Ray
  • Compute scales are listed from 411M to 7B parameters

FAQ

How much corpus is listed?

The material states more than 300T of unfiltered CommonCrawl data.

What scales are supported?

The material lists 411M to 7B parameter models.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review