Model components / Base / pretrained models

Marin

Stanford CRFM's Marin, a fully-open 'open lab' foundation model (Marin 8B Base/Instruct and 32B Base), built in JAX/Levanter with bit-for-bit reproducibility. Code, data, experiments, hyperparameters and training logs are all documented and released.

Apache-2.0; ~12T-token training run with full data/code/logs and bit-reproducible training. Fully open but below OLMo 3 on fully-open evals; modest adoption.

Openness

5 high confidence

5.0

weights: open(Apache-2.0)
data: open(documented mixtures)
code: open(marin + Levanter, JAX)
checkpoints: open
reproducibility: bit-for-bit
license: Apache-2.0(OSI)

Stanford CRFM 'open lab': Apache-2.0 weights plus documented code, data, experiments, hyperparameters and training logs, with bit-for-bit reproducible training.

http://marin.community/blog/2025/05/19/announcement/ recorded 2026-06-30
Marin open lab: code, data, experiments, logs all shared
https://huggingface.co/marin-community/marin-8b-base recorded 2026-06-30
Marin 8B Base on HF; marin-community

Adoption

2 low confidence

2.0

Research-stage fully-open project; modest adoption. Included as a fully-open exemplar and to track Stanford CRFM's entry, not on popularity.

https://huggingface.co/marin-community/marin-8b-base recorded 2026-06-30
marin-community HF org; research-stage uptake

Capability

3 medium confidence

3.0

Mid-tier fully-open capability at 8B/32B; OLMo 3 outperforms it on fully-open model evaluations.

https://developers.googleblog.com/stanfords-marin-foundation-model-first-fully-open-model-developed-using-jax/ recorded 2026-06-30
Marin 8B fully-open, JAX/Levanter, 12T tokens

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