Find open-source science resources

Open-source scientific multimodal foundation model built on a 235B MoE LLM and 6B vision encoder, continually pretrained on 5T tokens including 2.5T scientific-domain tokens, with strong results across chemistry, materials, life science, and earth science benchmarks (2025)

High-accuracy PDF→Markdown/JSON/HTML conversion, specialized for tables/formulas/code blocks with benchmark scripts

Ontology representation of the [International Committee on Taxonomy of Viruses (ICTV)](https://ictv.global/) for the [EVORA project](https://evora-project.eu/)

Deterministic, rule-based variant interpretation platform for clinical genetics laboratories. Automates ACMG/AMP 2015 classification using a Bayesian point-based framework (Tavtigian et al. 2018) with BayesDel ClinGen SVI-calibrated thresholds (Pejaver et al. 2022). Integrates 8 reference databases (gnomAD v4.1, ClinVar, dbNSFP 4.9c, SpliceAI, gnomAD Constraint, HPO, ClinGen, Ensembl VEP). Analyzes nuclear and mtDNA variants, structural and copy-number variants (SV/CNV), with trio/family and cohort analysis. Supports HPO-based phenotype matching, biomedical literature mining across 2M+ PubMed publications, and structured clinical report generation. AI assists in evidence synthesis but does not make classification decisions. EU-hosted on dedicated infrastructure in Helsinki, Finland (GDPR-compliant).

Fully open-source (Apache 2.0) biomolecular structure prediction reproducing AlphaFold3, free for academic and commercial use (Columbia AlQuraishi Lab & OpenFold Consortium, 2025)

Google Research's hybrid ML/physics atmospheric model combining learned dynamics with physical constraints, outperforming traditional models on 2-15 day forecasts and 40-year climate simulation, developed with ECMWF (Nature 2024)

- Molecular Manipulation Made Easy. A light wrapper build on top of RDKit.

Terminal AI coding assistant with a built-in math formalization engine that converts plain-language math problems into Lean 4 theorems and attempts formal proofs; bundles a local Lean toolchain and WebUI for interactive mathematical reasoning (math-ai-org, 582+ stars, 2026)

Developer toolkit for accelerating training and inference for AI in chemistry and material science, providing optimized GPU-accelerated workflows for molecular and materials machine learning (NVIDIA, 2026)

Universal machine learning interatomic potential for atomistic simulation of materials, molecules, and biomolecules across the periodic table, with open-source pretrained models and inference tools (Orbital Materials, 2024-2025)

ECMWF's open-source machine-learning Earth system model developed by the WeatherGenerator Consortium with NVIDIA, trained on reanalyses, forecast data, and diverse observations across atmosphere, ocean, and land to provide a robust multi-scale model of Earth system dynamics; the first released version (v0.1, trained on ERA5) demonstrates global probabilistic forecasting skill on par with established AI models, with open training framework and config-driven multi-dataset ingestion pipeline (Apache 2.0)

Auto-generates clean, customizable academic CVs from open research data (OpenAlex, ORCID, Crossref, DataCite, Open Editors Plus). A single canonical CV object drives every output format (HTML, PDF, DOCX, LaTeX, Markdown); citations render through CSL; and the account holder is matched by persistent identifier (ORCID / OpenAlex ID) rather than name string. Free for individuals, open-source, and FAIR by design.

Generalist autonomous research agent that grows a hypothesis tree to optimize any measurable task, beating Claude Code and Codex by 2.5× on the same compute budget across BrowseComp, Terminal-Bench 2.0, math reasoning, and MLE-Bench Lite; supports native CLI, keyless Claude Code/Codex integration, and an MCP tool server (RUC-NLPIR, 866+ stars, Apache 2.0, 2026)

High-performance symbolic regression for discovering interpretable scientific equations from data, multi-population evolutionary search with Python/Julia backend, widely used in physics and astronomy (Cambridge, NeurIPS 2023)

Language agent gymnasium for challenging scientific tasks including DNA manipulation, literature search, and protein engineering

Democratizing AI scientists by transforming any LLM into research systems with 600+ scientific tools (Harvard MIMS)

Fits second-order autoregressive AR(2) models to gene expression time series and reports the eigenvalue modulus |lambda|, a single statistic quantifying temporal persistence: how strongly a gene's recent past constrains its next value. Ranks genes into a clock/target/background hierarchy and reports correlation length, half-life and root type (real or complex) per gene.

Robust, lightweight infrastructure for multi-agent autonomous self-evolution, built for autoresearch; agents run in isolated git worktrees, share knowledge through a common state directory, and are scored by a grader daemon; natively integrated with Claude Code, Codex, Cursor Agent, OpenCode, and Kiro (672+ stars, Apache 2.0)

Principle-first scientific idea discovery framework that extracts reusable principles from public literature and private research materials, composes them into traceable Idea Cards with prior-art comparisons, and exports validation-ready research packs; emphasizes inspectable scientific objects, risk disclosure, and falsification paths (ICML 2026, 411+ stars, MIT License)

Benchmark evaluating AI agents on complex real-world scientific workflows in terminal environments across life, physical, earth, and mathematical sciences; featured on model cards for Claude Opus 4.7, GPT-5.5, and Gemini 3.1 Pro (200+ stars, Apache 2.0)

High-performance ML research

NVIDIA and King's College London's open-source AI toolkit for healthcare imaging, providing foundational frameworks for medical image annotation (MONAI Label), training (MONAI Core), and deployment (MONAI Deploy) across radiology, pathology, and endoscopy (8K+ stars, Apache 2.0)

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

ReviewAid is an open-source AI-assisted tool for full-text screening and data extraction in systematic reviews. It supports evidence synthesis workflows by using large language models to classify articles according to user-defined PICO criteria and extract structured information from full-text publications. ReviewAid is designed as a supplementary reviewer rather than a replacement for human judgement. It aims to reduce manual workload, improve consistency, and assist researchers during screening and data extraction while maintaining human oversight throughout the evidence synthesis process.

Family of operational-quality open weather models from DeepMind and Google Research, including WeatherNext Graph (deterministic GNN medium-range forecasting, published as GraphCast), WeatherNext Gen (diffusion ensemble, published as GenCast), WeatherNext 2 (state-of-the-art global medium-range and cyclone forecasting skillful beyond 15 days, operational at 0.25° resolution), and WeatherNext Cyclones (breakthrough tropical cyclone track forecasting, Nature 2026); official code and open weights (Apache 2.0, 7.6K+ stars)

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)

Noodle is a public, read-only web application for semantic discovery and exploration of biomedical literature sourced from PubMed, Europe PMC, and OpenAlex. It accepts natural-language research questions and presents ranked publications together with gene, genetic variant, and phenotype views. Publication pages provide provenance, identifiers, source links, indexed concepts, and navigable semantic-neighbourhood graphs. The companion Noodle Biomedical Literature Discovery MCP exposes the same public discovery domain through a read-only, source-linked Model Context Protocol interface with bounded citation and semantic graph traversal. Semantic proximity, citation distance, co-mention, and search rank are discovery signals and do not establish scientific evidence, causality, diagnosis, or treatment.

xgt is a command-line tool for programmatic access to the GTDB REST API. It provides four subcommands: search (genome queries with pagination), genome (cards, metadata, taxonomic history), taxon (lineage and genome set retrieval), and diff (per-rank taxonomic comparison between any two GTDB releases). All subcommands support batch input, JSON/CSV/TSV output, file splitting, and automatic retry. Implemented in Rust as a self-contained binary with no runtime dependencies.

Inflexa is an open-source, agentic orchestration platform for computational biology and translational medicine. It is designed to assist researchers in analyzing multi-omics, cheminformatics, and imaging data by reading published literature, designing multi-step analysis plans, and executing experiments with full reproducibility.

Free-text promptable universal 3D medical image segmentation foundation model enabling zero-shot segmentation of diverse anatomical structures and pathologies via natural language prompts across CT, MRI, and other volumetric imaging modalities (DKFZ, 195+ stars, Apache 2.0)

Self-evolving AI scientist with 6 specialized sub-agents (plan/research/code/debug/analyze/write) and persistent memory, #1 on DeepResearch Bench II and AstaBench, supporting multi-provider LLMs and multi-channel deployment (Apache 2.0, 2026)

FAIR EVA: Evaluator, Validator & Advisor has been developed to check the FAIRness level of digital objects from different repositories or data portals. It requires the object identifier (preferably persistent and unique identifier) and the repository to check. It also provides a generic and agnostic way to check digital objects. This software started to be developed within IFCA-Advanced-Computing receives funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 857647.

High-accuracy RAG for scientific PDFs with citation support, agentic RAG, and contradiction detection

World's first fully open, accelerated weather AI software stack with Medium Range forecasting and Nowcasting models using generative AI (January 2026)

Apache 2.0 single-cell foundation model family scaling to 3B parameters, pretrained on 266M cell profiles including perturbation data and released with training, embedding, and downstream benchmarking workflows for disease-relevant single-cell tasks (2025)

SQUARNA is a tool for RNA secondary structure prediction. It can take a single RNA sequence or an alignment of sequences as input. SQUARNA handles pseudoknots and can predict alternative structures. SQUARNA allows structural restraints and chemical probing data as additional input and is available at https://github.com/febos/SQUARNA and https://larnal.imol.institute/.

General-purpose biomedical AI agent integrating LLM reasoning with retrieval-augmented planning and code-based execution to autonomously execute diverse biomedical research tasks and generate testable hypotheses (Stanford SNAP, bioRxiv 2025)

A structural pharmacology workbench for cognition-related CNS targets, built so that a displayed number must trace to a computation. Runs Boltz-2 locally for structure prediction, validates chemistry with RDKit, enforces a provenance record on every value, and reports eight studies pre-registered under content hashes before any data was seen. Its headline result is negative: designed peptides did not separate from composition-matched shuffles of their own amino acids.

The submission-centric metadata schema for the German Human Genome-Phenome Archive (GHGA).

Python package for simulation-based inference enabling likelihood-free Bayesian parameter estimation from scientific simulators, with flexible interfaces for neural posterior estimation, sequential methods, and MCMC/variational backends (Mackelab, 825+ stars)

Scikit-learn compatible tabular foundation model for zero-shot classification and regression on mixed-type tabular datasets via in-context learning; applicable to diverse scientific datasets (1.8K+ stars, Apache 2.0)

Molecular dynamics in JAX

NOS-TLPlot is an open-source tool for visualizing Newcastle–Ottawa Scale (NOS) risk-of-bias assessments in systematic reviews. It converts NOS star ratings into publication-ready traffic-light plots and 12 specialized visualizations, enabling reviewers and readers to interpret study-level risk-of-bias results clearly and reproducibly.

Knowledge graph-guided synthetic data generation for LLM fine-tuning, achieving strong performance on scientific QA (GPQA-Diamond) and math reasoning (AIME)

Open-source, all-atom biomolecular foundation model that turns co-folding into a scalable engine for structure prediction, design, and optimization across proteins, nucleic acids, and small molecules in drug discovery; ranked first on PXMeter-AB, FoldBench-AB, and 2026ARK-AB antibody-antigen benchmarks (263+ stars, Apache 2.0)

A quantum chemistry package written in Python.

a specification for describing analysis workflows and tools that are portable and scalable across a variety of software and hardware environments, from workstations to cluster, cloud, and high performance computing (HPC) environments.

Pretrained time series foundation model for zero-shot forecasting across diverse scientific and real-world domains; tokenizes continuous time series into discrete bins to train transformer language models on large-scale corpora, achieving strong zero-shot generalization and competitive performance with task-specific supervised models on climate, energy, and health benchmarks (5.3K+ stars, Apache 2.0, 2024-2026)

Google DeepMind's diffusion-based ensemble weather forecasting model at 0.25° resolution, outperforming ECMWF ENS on 97.2% of targets up to 15 days ahead, with open-source code and weights (Nature 2024)