Find open-source science resources

The MediPhi Model Collection comprises 7 small language models of 3.8B parameters from the base model Phi-3.5-mini-instruct specialized in the medical and clinical domains. The collection is designed in a modular fashion. Five MediPhi experts are fine-tuned on various medical corpora (i.e.

## Description: Geneformer is a foundational transformer model pretrained on a large-scale corpus of single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology.

## Description: Geneformer is a foundational transformer model pretrained on a large-scale corpus of single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology. This model version was continually pretrained on ~14 million cancer transcriptomes…

## Description: Geneformer is a foundational transformer model pretrained on a large-scale corpus of single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology.

Python-centric Cookiecutter for Molecular Computational Chemistry Packages by [MolSSL](https://molssi.org/)

## Description: Geneformer is a foundational transformer model pretrained on a large-scale corpus of single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology.

Multimodal whole-slide pathology foundation model jointly pretrained on H&E histology and diagnostic text reports, enabling zero-shot cancer subtyping, biomarker prediction, and multimodal reasoning across diverse cancer types (Mahmood Lab, 341+ stars)

For a convenient overview and download list, visit our model page for this model.

Support our open-source dataset and model releases!

Full spaCy pipeline and models for scientific/biomedical documents, enabling named entity recognition, abbreviation resolution, and UMLS linking for scientific literature mining (1.9K+ stars, Apache 2.0)

The Bibframe vocabulary consists of RDF classes and properties used for the description of items cataloged principally by libraries, but may also be used to describe items cataloged by museums and archives. Classes include the three core classes - Work, Instance, and Item - in addition to many more classes to support description. Properties describe characteristics of the resource being described as well as relationships among resources. For example: one Work might be a "translation of" another Work; an Instance may be an "instance of" a particular Bibframe Work. Other properties describe attributes of Works and Instances. For example: the Bibframe property "subject" expresses an important attribute of a Work (what the Work is about), and the property "extent" (e.g. number of pages) expresses an attribute of an Instance.

Autonomous multi-agent research loop for model architecture discovery that ran 1,773 experiments over 20,000 GPU hours and produced 106 state-of-the-art linear-attention architectures, surpassing human-designed baselines including Mamba2 and DeltaNet (1.1K+ stars, Apache 2.0)

Hulu-Med: A Transparent Generalist Model towards Holistic Medical Vision-Language Understanding

Large Language and Vision Assistant for bioMedicine (i.e., “LLaVA-Med”) is a large language and vision model trained using a curriculum learning method for adapting LLaVA to the biomedical domain. It is an open-source release intended for research use only to facilitate reproducibility of the…

Tool designed to provide a simple way of standardising molecules as a prelude to e.g. molecular modelling exercises.

A crystallography domain ontology based on EMMO and the CIF core dictionary. It is implemented as a formal language. (from https://nfdi4cat.org/services/ontologie-sammlung/)

A vocabulary describing licenses for data and code

Autonomous algorithm discovery combining evolutionary search with peer-review reward models, achieving best-known performance on circle packing problems

!1

BulkRNABert is a transformer-based, encoder-only language model pre-trained on bulk RNA-seq profiles from the TCGA dataset using self-supervised masked language modeling, following the original BERT framework. The model is trained to reconstruct randomly masked gene expression values from their…

Graph neural network operating entirely at the atomic level for protein-ligand conformational ensemble prediction and docking, generating diverse solutions through rapid stochastic denoising to model conformational heterogeneity (Baker Lab, bioRxiv 2025)

ChemFIE-BED is a sentence-transformers based on gbyuvd/chemselfies-base-bertmlm fine-tuned on around (for now) 2 million pairs of valid molecules' SELFIES (Krenn et al. 2020) taken from COCONUTDB (Sorokina et al. 2021) and ChemBL34 (Zdrazil et al. 2023).

Family of codon-resolution language models trained on 130 million protein-coding sequences from over 20,000 species, enabling cross-species gene expression prediction and codon-level functional genomics (2025)

ChemFormula provides a class for working with chemical formulas. It allows parsing chemical formulas, calculating formula weights, and generating formatted output strings (e.g. in HTML, LaTeX, or Unicode).

GitHub homepage: Cell2Sentence GitHub

Discovering interpretable features in protein language models via sparse autoencoders, enabling mechanistic understanding of PLM representations for protein engineering and design (288+ stars, MIT License)

First versatile medical reasoning agent for chest X-ray interpretation, dynamically integrating state-of-the-art CXR analysis tools and multimodal LLMs into a unified framework; introduces ChestAgentBench with 2,500 complex medical queries across 7 categories (bowang-lab, 1.1K+ stars)

A library for computational chemistry (DFT) for input file generation, data extraction, method screening and analysis.

S3segmenter is a Matlab-based set of functions that generates single cell (nuclei and cytoplasm) label masks.

Experiments with expanded ensembles to explore chemical space.

Conversational data analysis using natural language

This model is a fine-tuned version of google/medgemma-4b-it adapted for binary mammogram classification on the OMAMA 256×256 dataset. The dataset consists of ~154k mammogram image slices (.npz) with metadata JSONs providing labels (NonCancer, Cancer).

For a convenient overview and download list, visit our model page for this model.

AI-powered pipeline converting papers into interactive websites, posters, and multimedia presentations with "Let's Make Your Paper Alive!" philosophy

Geometry Aware Operator Transformer serving as an efficient and accurate neural surrogate for PDEs on arbitrary domains, combining geometric priors with transformer architectures for scientific computing (ETH Zurich CAMLab, 92+ stars)

The Nucleotide Transformers are a collection of foundational language models that were pre-trained on DNA sequences from whole-genomes. Compared to other approaches, our models do not only integrate information from single reference genomes, but leverage DNA sequences from over 3,200 diverse human…

A Package For Training SNAP Interatomic Potentials for use in the LAMMPS molecular dynamics package.

Autonomous pipeline from literature review→hypothesis→algorithm implementation→publication-level writing with Scientist-Bench evaluation

DermLIP is a vision-language model for dermatology, trained on the Derm1M dataset—the largest dermatological image-text corpus to date. This model variant (PanDerm-base-w-PubMed-256) utilizes domain-specific pretraining to deliver superior performance compared to other DermLIP variants..

Generalist foundation model and database for open-world medical image segmentation, enabling universal segmentation of diverse anatomical structures and pathologies with zero-shot generalization to unseen tasks and modalities (Nature Biomedical Engineering 2025)

MedVAL-4B (medical text validator) is a language model fine-tuned to assess AI-generated medical text outputs at near physician-level reliability.

LLM agent system synthesizing Wikipedia-like long-form research articles from scratch through multi-perspective question asking, web retrieval, and citation-grounded report generation, with Co-STORM extension for collaborative human-LLM knowledge curation conversations (Stanford OVAL, NAACL 2024 & EMNLP 2024)

Automated and rigorous experiments using AI agents for scientific discovery

This model is a lightweight model pre-trained on SELFIES (Self-Referencing Embedded Strings) representations of molecules. It is trained on 2.7M unique and valid molecules taken from COCONUTDB and ChemBL34, with 7.3M total generated masked examples.

> [!NOTE] > This model has been optimized using NVIDIA's TransformerEngine > library. Slight numerical differences may be observed between the original model and the optimized > model. For instructions on how to install TransformerEngine, please refer to the > official documentation.

> [!NOTE] > This model has been optimized using NVIDIA's TransformerEngine > library. Slight numerical differences may be observed between the original model and the optimized > model. For instructions on how to install TransformerEngine, please refer to the > official documentation.

Partially latent flow matching model for the joint generation of a protein's amino acid sequence and full atomistic structure, including both backbone and side chains (2025)

Semantic-enhanced multi-modal remote sensing foundation model for Earth observation (Nature Machine Intelligence 2025), enabling universal interpretation across diverse satellite imagery modalities with open-source weights and benchmarks

Website    🤖 7B Model    🤖 32B Model    MedEvalKit    Technical Report    Lingshu MCP