AI Potluck
Model components / Dataset Processing Tools

Marin (framework)

Stanford CRFM

Open framework for reproducible foundation-model development that tracks every step from raw data to final model: data curation, transformation, filtering, tokenization, training, evaluation, and experiment/mixture orchestration. Distinct from the Marin model it produces.

Stanford CRFM / Marin community framework (marin-community/marin), distinct from the Marin model artifact on HF; repo very active (~9.5k commits), Apache-2.0 confirmed live 2026-07-21. Contains no weights; produces the Marin 8B model.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(marin-community/marin)
governance
Marin community/Stanford CRFM
core-gated
ungated

Fully OSI-licensed (Apache-2.0), full source public. Repo is the framework; the Marin model is a separate artifact.

Adoption

2 medium confidence
2.0

~1.2k GitHub stars and very active development; a research framework with adoption concentrated around the Marin project rather than broad external use.

Capability

4 medium confidence
4.0

Broad end-to-end framework; adoption, not capability, is its weak axis.

Unchanged since 2026-07-30 (last edited, not re-checked)