Methodology

What does prominence mean?

← Back to the map

Prominence is a proxy for community recognition and usage. Not commercial success, not clinical deployment scale, not a claim about safety.

A model with 2,000 citations and 50k monthly HuggingFace downloads from a top lab is clearly prominent. A single preprint with no external uptake is not, whatever its authors claim. The rubric below formalises this without pretending to be objective.

i.The rubric

Four signals, scored 0–2 each: provenance plus three usage signals. Only a model's two strongest usage signals count, so a channel nobody uses (a quiet repo, no paper yet) never counts against it. Click a row for the rationale behind the threshold.

Signal
Strong (2)
Moderate (1)
Weak (0)
GitHub stars
≥ 1,000
200–999
< 200
HF monthly downloads
≥ 10k
1k–10k
< 1k
Citations
≥ 1,000
200–999
< 200
Org / provenance
Named industry, biotech, or top academic lab
Named lab, weak traction
Unknown / solo
+1 Activity modifier. One extra point if there's evidence of commits, releases, or vendor announcements in the last 6 months. No penalty for inactivity. Settled-but-used models stay where they land.

Bands

High
≥ 5 of 7 pts
Org plus best two usage signals plus activity. Needs data on ≥ 2 usage signals. Marks a model prominent.
Medium
4 pts
Flagged for manual review before inclusion.
Low
< 4 pts
Skipped, with a one-line note in the proposals log.

Two structural fixes sit on top:

  1. Clinical-deployment substitute. Clinical FMs often have no repo or HF presence. Documented use with real patients or clinicians at a named health system (pilots included) counts as a usage signal worth 2 points.
  2. Recency carve-out. A model released in the last 12 months from a credentialed lab, with early traction (100+ citations, 200+ stars, or 1k monthly downloads), gets a one-band bump. Counters the structural penalty new releases face on stars, citations, and downloads.

ii.The three gates before any of that

The rubric only runs on candidates that pass all three:

  1. Must be a foundation model. Pretrained, general-purpose or multi-task substrate. Single-task classifiers are excluded.
  2. Must operate on HCLS data. DNA, RNA, protein, SMILES, whole-slide imaging, EHR text, medical imaging, surgical or endoscopy video, related modalities. General-purpose LLMs (GPT-4, Claude, Gemini) are excluded, even when used clinically.
  3. Must fit a category on the map. One of the map's categories, or Emerging if it is cross-cutting.

iii.Prominent vs tracked

Every entry carries one of two tiers:

Prominent
Clears the high band
Scores high on the rubric. The rule runs both ways: a tracked model that clears it is promoted, and a prominent one that falls below it is demoted.
Tracked
Worth knowing about
Cleared the gates, but not the high band. The default tier for new entries.

iv.What this map does not track

  1. Detailed license terms. The captures whether weights are downloadable. I don't track MIT vs Apache vs commercial beyond that.
  2. Clinical validation or regulatory approval.
  3. Safety, bias, or fairness evaluations.
  4. Commercial availability or pricing.
  5. Whether I agree with the model's framing or stated use case.

v.Where the numbers come from

Stars from the GitHub API, monthly downloads from the HuggingFace API (official repos only, all sizes summed), citations from Semantic Scholar or OpenAlex. Every score keeps its source links.

Earlier versions also scored cross-listing across five GitHub awesome-lists and five surveys. That signal was dropped in November 2026: the lists stopped updating in early 2025 and read more like personal bookmarks than evidence of use. New models are now found through the HuggingFace Hub.

Last updated: November 2026 Back to the map