Find open-source science resources

Open-source self-supervised vision foundation model for Earth observation by Clay Foundation (non-profit), a Masked Autoencoder ViT pretrained on multimodal satellite imagery (Sentinel-1/2, Landsat 8-9, NAIP, MODIS, LINZ DEM) with location/time embeddings, supporting classification, segmentation, change detection, similarity search, and few-shot downstream geospatial tasks (Apache 2.0, v1.5 2024-2025)

Fast, all-atom SE(3)-equivariant diffusion model for protein design achieving state-of-the-art performance on unconditional generation, motif scaffolding, and binder design while retaining the computational efficiency of equivariant architectures (bioRxiv 2026)

Medical large vision-language model unifying comprehension and generation via heterogeneous knowledge adaptation, enabling holistic medical image understanding, visual question answering, and clinical report generation across diverse modalities (ZJU4HealthCare, 1.6K+ stars)

A patient-level disease classification model trained on single-cell RNA-seq data. Given a matrix of gene expression profiles (one row per cell), the model produces a disease-category prediction for the patient.

MedPsy-4B is a state-of-the-art, text-only medical and healthcare language model purpose-built for edge deployment. Built on top of Qwen3-4B-Thinking-2507 and post-trained with a multi-stage pipeline (supervised fine-tuning + reinforcement learning) on curated medical data, it surpasses models…

MedPsy-1.7B is a state-of-the-art, text-only medical and healthcare language model purpose-built for edge and smartphone deployment. Built on top of Qwen3-1.7B (operated in thinking mode, i.e. with enable_thinking=True) and post-trained with a multi-stage pipeline (supervised fine-tuning +…

Trainable PyTorch reproduction of AlphaFold 3

Benchmark quantifying end-to-end autonomous AI research abilities of LLM agents across 20 tasks from SOTA machine learning papers spanning NLP, code, math, biochemical modelling, and time series forecasting, with normalized score metrics against human SOTA and HuggingFace dataset

General-purpose RNA language model with 650M parameters pretrained on 36M non-coding RNA sequences, achieving strong generalization on structure prediction tasks including secondary structure prediction, splice-site prediction, mean ribosome loading, and ncRNA classification (lbcb-sci, 165+ stars, Apache-2.0)

First physics-aligned interactive benchmark for LLM agents in engineering construction, designing rockets/cars/bridges in physics simulator with 3D spatial geometry library

Simple RDKit molecule editor GUI using PySide.

Semi-flexible molecular diffusion model for structure-based drug design with reinforcement learning, achieving 20× faster sampling and providing a no-code web platform for molecular design (ISPC Lab, Tongji University, 2026)

Incremental knowledge graph construction using LLMs with entity extraction and Neo4j visualization

Fast spike sorting with drift correction for extracellular electrophysiology, enabling universal neural spike sorting via deep learning on high-density neural probe recordings (MouseLand, 609+ stars)

The application of 2nd and 3rd generation High Throughput Sequencing (HTS) technologies has deeply reshaped experimental method to investigate microbial communities and obtain a taxonomic and functional profile of the invetigated community. Shotgun Metagenomics allow to quickly obtain a representation of microorganisms genomes characterizing a particular environment. In order to obtain a fast e reliable taxonomic classification of microorganisms genomes we present kMetaShot, an alignment-free taxonomic classifier based on k-mer/minimizer counting.

## Introduction (简介) This model is a domain-specific expert fine-tuned from Qwen/Qwen2.5-7B-Instruct using LoRA (Low-Rank Adaptation). It is specifically designed for Fine-grained Information Extraction (IE) of technical indicator quintuples from highly complex lithium-ion battery patents.

Trackastra is a transformer-based cell-tracking tool for live-cell microscopy. It links already segmented cell instances across time by predicting associations between detections. It supports greedy tracking with or without cell divisions, optional ILP-based linking, pretrained tracking models, and export of tracked masks and lineage information in Cell Tracking Challenge format.

Benchmark evaluating AI agents on 75 curated Kaggle-style ML engineering competitions with reproducible Docker-based grading harness, human baselines, and end-to-end task lifecycle, used as a primary benchmark for autonomous ML research agents (e.g., InternAgent #1 at 36.44%)

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

An extension of Schema.org to annotate metadata on software projects

Protein structure prediction

Distributional flow matching model for robust single-cell perturbation prediction, modeling the full distribution of perturbed cellular expression profiles conditioned on control states via PAD-Transformer and multi-kernel MMD regularization; reduces MSE by 19.6% over the strongest baseline in combinatorial settings (Westlake University, 41+ stars, MIT License)

FutureHouse's end-to-end scientific discovery multi-agent system orchestrating literature search (Crow/Falcon) and data analysis (Finch) agents, first AI-generated drug discovery identifying ripasudil as novel dry AMD therapeutic (2025)

Benchmark evaluating AI agents' ability to replicate 20 ICML 2024 Spotlight/Oral papers from scratch, with 8,316 gradable tasks and author-co-developed rubrics

End-to-end molecular dynamics engine built on PyTorch, enabling differentiable simulations with neural network potentials and GPU acceleration for machine learning-accelerated molecular dynamics (MIT License, 707+ stars)

Scaling efficient, expressive, and general SE(3)-equivariant graph attention transformers for atomic systems and machine-learned interatomic potentials (MIT License, 2026)

Pretrained machine-learned force field for (bio)molecular simulations combining the fast SO3krates neural network for semi-local interactions with universal pairwise force fields for short-range repulsion, long-range electrostatics, and dispersion interactions; supports geometry optimization, NVT/NPT/NVE MD, fine-tuning, ASE calculator, and JAX-MD integration (JACS 2025, 218+ stars, MIT License)

Duchifat-2.3-Instruct is a state-of-the-art, instruction-tuned Large Language Model developed by TopAI. As the flagship of the Duchifat series, this model represents a fundamental breakthrough in how Hebrew is processed, reasoned, and generated in the LLM era.

!image

Fully autonomous medical image segmentation research system that generates complete manuscripts end-to-end from datasets with zero human intervention, beating strongest baselines on 24 of 31 datasets and achieving T1-T2 tier manuscript quality in double-blind evaluations (USTC & Shanghai AI Lab, 2026)

Unified latent diffusion transformer that jointly generates periodic crystals and non-periodic molecules, scaling to 500M parameters with SOTA results on QM9, MP20, and GEOM-DRUGS (Meta FAIR, ICML 2025, 310+ stars)

The EMI ontology is used to structure spectrum annotation provenance by reusing the PROV-O ontology (a W3C recommendation) and sample and observation data by applying the SOSA ontology. EMI reuses the SOSA ontology as a data schema for struturing the Sample and Observation data. SOSA (Sensor, Observation, Sample, and Actuator) is a subset of SSN (Semantic Sensor Network Ontology) that is a W3C recommendation. [from homepage]

AstraZeneca's industrial-grade retrosynthetic planning tool using MCTS to recursively decompose molecules into purchasable precursors, with multi-step route scoring and support for custom one-step models (v4.0, 2024)

Fine-tuned version of google/gemma-4-E4B-it across three professional domains — Medical, Legal, and Finance — using QLoRA (4-bit NF4) with Optuna-tuned hyperparameters, trained on Kaggle T4 GPU.

Scalable agentic training environment for code-centric reasoning in biomedical data science

The Context and Measurement Ontology (COMO) contains ontological terms to describe the context for various types of experimental data and measurements. It is useful in its current state for several different environmental microbiology projects. This ontology is used in multiple CORAL (Contextual Ontology-based Repository Analysis Library) deployments.

L1 (Learning Unit 1) is the first language model from Lunit and Lunit Consortium, purpose-built for the medical domain. Derived from Gravity-16B-A3B-Base, L1 is designed for clinical reasoning and decision support.

Improved equivariant Transformer for 3D atomic graphs (ICLR2024)

DeepConsensus uses gap-aware sequence transformers to correct errors in Pacific Biosciences (PacBio) Circular Consensus Sequencing (CCS) data.

Accessible protein design platform via Google Colab integrating AlphaFold2, RoseTTAFold, and ProteinMPNN for de novo hallucination, fixed backbone design, and binder design (Sergey Ovchinnikov, 2022+)

Extended autonomy AI scientist with 200 parallel agent rollouts, 42K lines of code execution, 1.5K papers analyzed per run, achieving 79.4% accuracy and 7 scientific discoveries (Edison Scientific)

## Model Description This is a lightweight, high-performance image classification model built to diagnose histopathological scans of lung and colon tissues. This model was specifically designed for rapid web deployment without sacrificing clinical accuracy.

Provides R wrappers of several on-target and off-target scoring methods for CRISPR guide RNAs (gRNAs). The following nucleases are supported: SpCas9, AsCas12a, enAsCas12a, and RfxCas13d (CasRx). The available on-target cutting efficiency scoring methods are RuleSet1, RuleSet3, DeepHF, enPAM+GB, and CRISPRscan. Both the CFD and MIT scoring methods are available for off-target specificity prediction. The package also provides a Lindel-derived score to predict the probability of a gRNA to produce indels inducing a frameshift for the Cas9 nuclease. Note that DeepHF and enPAM+GB are not available on Windows machines.

First large vision-language assistant for gigapixel whole-slide pathology image understanding, released with the SlideInstruction dataset and SlideBench benchmark (uni-medical, Apache 2.0, 2025)

LLM agent framework for Earth Observation with 104 specialized tools across 5 functional kits