Find open-source science resources

💻 Github | 📄 Report (Coming soon...) | 🚀 Demo

!image

A 1.7B-parameter causal language model distilled from Qwen3-30B-A3B on 6,122 STEM chain-of-thought samples using discrepancy-informed knowledge distillation. The training objective emphasizes proof structure, detects reasoning pivot tokens through token-level divergence dynamics, smooths…

Fine-tuned ESM-2 650M with LoRA for predicting protein subcellular localization (10 classes).

PhenoVisionL is a Vision Transformer (ViT-Large) model fine-tuned to detect leaf phenological states in plant photographs: green leaves, colored (senescent) leaves, and breaking leaf buds. It was trained on 165,988 iNaturalist records of deciduous woody plants using a two-stage semi-supervised…

PhenoVision is a Vision Transformer (ViT-Large) model fine-tuned to detect flowers and fruits in plant photographs. It was trained on 1.5 million human-annotated iNaturalist images and has been used to generate over 30 million new phenology records across 119,000+ plant species, vastly expanding…

> NEXUS domain specialist for medical Q&A and clinical reasoning — lightweight & uncensored.

> NEXUS domain specialist for medical Q&A and clinical reasoning — lightweight & uncensored.

Distilling Boltz: Flow Maps for Fast All-Atom Cofolding

!Screenshot 2026-07-05 at 2.33.47 AM

### Model Overview TabPFN-3 is a transformer-based foundation model that uses in-context-learning to solve tabular prediction problems in a forward pass. Inference code can be found at https://github.com/PriorLabs/TabPFN. More details can be found in the Model Report.

This repository contains LoRA finetunes of DiffusionGemma (image-conditioned discrete-diffusion LLM) for radiology visual question answering, each paired with an autoregressive Gemma-4 finetune as a controlled baseline. It corresponds to the paper Discrete Diffusion Language Models for Interactive…

Fine-tuned gSelformer-MV-Dyn model for predicting dynamic surface tension (DST) curves from SMILES, temperature, and concentration.

A native MLX port of OpenMed/privacy-filter-multilingual-v2 for Apple Silicon PII detection and de-identification with OpenMed. This is the unquantized BF16 reference artifact. For the 8-bit sibling, see OpenMed/privacy-filter-multilingual-v2-mlx-8bit.

A native MLX port of OpenMed/privacy-filter-multilingual-v2, affine-quantized to 8-bit for faster and smaller Apple Silicon PII detection with OpenMed. For the unquantized BF16 reference, see OpenMed/privacy-filter-multilingual-v2-mlx.

!ESMC ProtHash Banner

!Python

🤗 Blog | 📄 Paper | 💻 Code | 🌐 FineMed | 🩺 DoctoBERT

🤗 Blog | 📄 Paper | 💻 Code | 🌐 FineMed | 🩺 DoctoBERT

Heretic-abliterated version of Qwen/Qwen2.5-0.5B-Instruct for the Evolva drug discovery pipeline.

HantaBERT fine-tunes DNABERT-2 on hantavirus RNA sequences for three simultaneous classification tasks: species/lineage, host, and geographic origin. A single forward pass produces predictions for all three tasks along with a 768-dimensional embedding suitable for phylogenetic visualization.

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

Xinghe1-9B (杏核) is a specialized large language model fine-tuned for the formalization, computational derivation, and clinical reasoning of Huangdi Neijing. It is based on the Qwen3.5-9B-Instruct architecture and trained using the V3 Double-Purity SFT dataset.

Pretrained LDARNet (~2M params) with learnable DNA tokenization (dynamic chunking + BiMamba-2).

Pretrained LDARNet (~110M params) with learnable DNA tokenization (dynamic chunking + BiMamba-2).

# Overview This is the CellHermes model, based on the LLaMA-3.1-8B-instruct architecture developed by Meta, fine-tuned using single-cell RNA sequencing (scRNA-seq) datasets from CellxGene and PPI network from BioGRID. CellHermes is an innovative framework for adapting existing large language models…

This repository contains Chemistry fine-tuned Qwen3-4B checkpoints from the local SciKnowEval-style generalization setup.

PlantGeneAnn is a plant genome foundation model that enables the prediction of various plant genomic elements at single-nucleotide resolution. The model is built upon the PlantBiMoE architecture with a 1D U-Net segmentation head, specifically designed for automated plant genome annotation.

33 lightweight, standalone models that predict discrete Alzheimer's-disease–related phenotypes (e.g. medication use, APOE genotype, vascular pathology, sex) from SomaScan plasma proteomics. The full list is in phenotypes.tsv.

69 lightweight, standalone models that predict continuous Alzheimer's-disease–related phenotypes (cognition, neuropathology burden, motor/functional measures, demographics, a genetic risk score, and longitudinal change) from SomaScan plasma proteomics.

This repository contains a drop-in, Hugging Face–compatible checkpoint converted from https://huggingface.co/microsoft/llava-med-v1.5-mistral-7b. You can load it with the exact same code you use for the original model—no extra conversion steps required.

👋 Join our LiGHT community. 📖 Check out the MeditronFO blog and MeditronFO preprint. 🔜 If you are a clinician join the MOOVE initiative here.

Sexo-FR is a French-language conversational language model that provides reliable, caring, and evidence-based sexual health information (information en santé sexuelle). It is part of a French public-health initiative whose goal is to make trustworthy sexual-health information more accessible to the…

👋 Join our LiGHT community. 📖 Check out the MeditronFO blog and MeditronFO preprint. 🔜 If you are a clinician join the MOOVE initiative here.

A lightweight plasma-protein aging clock that predicts chronological age from 85 unique inflammatory proteins measured by SomaScan (125 aptamers / SomaScan features). The model is a TabM† student distilled from a TabPFN v2 teacher, so it runs at inference without any TabPFN dependency (small,…

A lightweight plasma-protein aging clock that predicts chronological age from 92 Olink Inflammation-panel proteins. The model is a TabM† student distilled from a TabPFN v2 teacher, so it runs at inference without any TabPFN dependency (small, DUA-friendly artifacts).

Jolia is a 3D CT foundation model that encodes images into vector representations program. It encodes a whole 3D CT volume into:

Molexar-10M Base is the unconditional base model for Molexar, a unified multimodal molecular foundation model for drug design. It is trained as an autoregressive molecular language model over Fragment-SELFIES, a BRICS-fragment molecular language with validity-preserving decoding and…

Molexar-10M Omni is the universal multi-condition model for Molexar, a unified multimodal molecular foundation model for drug design. It starts from fairydance/molexar-10m-base and is supervised fine-tuned to generate Fragment-SELFIES molecules under scalar molecular-property,…

JEPA-based spatial-transcriptomics foundation model (TERRA). Code & docs: https://github.com/Lotfollahi-lab/terra

GRamma-12B is a 12-billion-parameter instruction-tuned language model specialized for the Greek medical domain. It is built on top of Gemma 3 12B Instruct and adapted through parameter-efficient fine-tuning on a collection of Greek and bilingual medical question-answering data.

Full weight-level fine-tuning of InstaDeepAI/nucleotide-transformer-v2-50m-multi-species for binary DNA sequence classification on two GenomicBenchmarks tasks. All parameters are updated rather than using LoRA or a frozen backbone, with a leakage-free train/validation/test protocol and multi-seed…

> Source code, training scripts, and inference utilities for this model: > github.com/NVIDIA-BioNeMo/KERMT > (v2.0 branch / v2.0.0 release tag)