Flagship resources by domain

Sensible "if in doubt, start here" picks per scientific domain. These are the resources users most often want when they describe a task in plain language. Always confirm with the live catalog (fetch_catalog.py topic <slug>) before final recommendation — the catalog evolves and there may be something newer.

The catalog is a discovery index; author cards, released code and live service metadata determine the usable interface. Reviewed 2026-10-01; no pretrained inference was run.

Biology / Genomics

DNA foundation models

• arcinstitute/evo2_40b — 40B DNA language model trained using the 8.8T-base OpenGenome2 corpus. Uses native Evo2/Vortex; current Hub provider mapping is empty. Arc documents NVIDIA hosting and multi-H100 local execution.
• arcinstitute/evo2_7b — 7B DNA language model, not an instruction-tuned chat variant. Native runtime supports BF16 without Transformer Engine; hardware feasibility depends on context and batch size.

Protein language models

• facebook/esm2_t33_650M_UR50D — 650M masked protein language model for embeddings or task-specific scoring. Its encoder alone does not output folded coordinates; folding requires a suitable model/head. Current HF provider mapping exposes fill-mask.
• Smaller ESM2 variants (*_t30_150M_*, *_t12_35M_*) for laptop/CPU use.

Single-cell / transcriptomics

• arcinstitute/Stack-Large (STACK) — single-cell foundation model with in-context learning across cell types.
• Merck/TEDDY — single-cell model/code repository trained on approximately 116 million cells; 116M describes training cells, not parameter count. Use its AnnData preprocessing and model classes.

Antibodies

• opig/OAS — Observed Antibody Space; inspect its paired/default configs and metadata before selecting a bounded subset.

Bioacoustics / ecology

• EarthSpeciesProject/NatureLM-audio — audio-language model for animal vocalizations; follow its dedicated loader and preprocessing.
• EarthSpeciesProject/esp-aves2-sl-beats-all — bioacoustic encoder requiring the AVEX library; verify the selected checkpoint's training and input contract.

Genomic corpora

• arcinstitute/opengenome2 — curated prokaryotic + eukaryotic sequences for foundation-model pretraining.

Chemistry / Drug discovery

• SandboxAQ/AQAffinity — gated drug-target affinity resource. Public metadata exists; card/artifact access requires accepted terms, so execution was not verified.
• SAIR dataset — protein-ligand structural/affinity resource (see SandboxAQ catalog entry). Its card was not accessible anonymously during review; verify source provenance, labels and terms before treating counts or structures as experimental ground truth.
• For SMILES/molecular tasks: check the catalog chemistry topic for current best — molecular foundation models change frequently.

Materials science

• LeMaterial — large open materials database (see catalog blog post 2024-12-10).
• Crystal structure foundation models, perovskite datasets — fetch topic materials-science for the live list. Many ship pymatgen.Structure objects rather than tensors.

Physics

• PDE-solver datasets (topic physics --filter datasets) — magnetohydrodynamics, fluid dynamics, plasma.
• Physics-Informed Neural Networks (PINN) — methodology covered in catalog blog posts.

Climate / Earth science / Weather

• Weather-foundation-model entries under topic climate — these are often huge multi-modal models with structured atmospheric inputs (geopotential, temperature on multiple pressure levels).
• Satellite imagery / remote sensing models under topic earth-science.

Medicine / Pathology

• hugging-science/breast-cancer-detector-2 — community breast ultrasound image-classification checkpoint. Catalog membership is not clinical validation; check cohort, patient splits, labels and intended use before benchmarking.
• Pathology + radiology foundation models — fetch topic medicine. Many are gated; check the model card for access requirements.

Mathematics / Scientific reasoning

• Kimina-Prover family — large theorem-proving models. Read the selected model card and verify a live provider mapping; availability is not implied by parameter count. Validate generated proofs with the exact Lean/mathlib environment.
• topic scientific-reasoning for LLM-as-scientific-assistant work, paper QA, multi-step reasoning evals.

Astronomy

• Galaxy survey datasets (e.g., hugging-science/mmu_legacysurvey_dr10_south_21) — HATS/Parquet release with image-array fields; inspect its schema and HATS reader requirements rather than assuming FITS files.
• Astronomical foundation models — fetch topic astronomy.

Cross-domain interactive demos (Spaces)

• hugging-science/BoltzGen_Demo — protein/peptide/nanobody binder design (source contract reviewed; runtime returned 503).
• hugging-science/dataset-quest — discover and submit scientific datasets.
• hugging-science/science-release-map — visualize who's publishing AI4Science resources.

How to use this list

1. User describes a task → match to a domain row above.
2. Fetch the live topic file with fetch_catalog.py topic <slug> and confirm the recommended resource still exists / is current.
3. Read the resource's HF card for input format, license, access requirements.
4. Follow using-datasets.md / using-models.md / using-spaces.md for the actual code.
5. If a related blog post is listed in the catalog, cite it when explaining methodology.

If nothing in this cheatsheet fits, run fetch_catalog.py search "<keyword>" against the full index. The catalog has hundreds of entries this file doesn't enumerate.

Primary cards: Evo2, TEDDY, STACK, AVEX checkpoint, Legacy Survey release.