RedPajama
Dataset
Curated Corpus
The open reproduction of Llama's training mix.
You're looking at a dataset record: a training-data collection — what's in it and which models learned from it.
Together AI and partners rebuilt Meta's disclosed Llama data recipe in the open — 1.2T tokens in v1, 30T documented in v2 — giving the open ecosystem a transparent, auditable base corpus.
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Related
Sources
Record reviewed as of 2026-01-15 — see Methodology.