Observed arrival · 2026-09-25
A family tree for machine-learning models—with crossed branches
Stemma Machinarum maps the lineages of open-weight language models and datasets, attaching evidence labels to each relationship.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Researchers tracing open-weight model lineage
- Worth noticing
- Edges are labeled as declared, uploader-declared, inferred, or alleged; cross-lineage influence is framed as “contamination.”
Field notes
The graph currently reports 96 models, 33 datasets, and 154 edges, with counts generated from the data at build time; its last data change is listed as 2026-09-24. Each edge carries an evidence label—such as declared, inferred from weights or behavior, or alleged—and the project provides a review procedure, disputes policy, raw data, and schema. The accompanying notebook is organized as six reading stations with dated notes and hands-on exercises.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue