The Best 92 Molecular Model Tools in 2025

A Transformer model based on the RoBERTa architecture, specifically designed for masked language modeling tasks with chemical SMILES strings

MoLFormer is a chemical language model pre-trained on 1.1 billion molecular SMILES strings from ZINC and PubChem. This version uses 10% samples from each dataset for training.

Evo is a biological foundation model capable of long context modeling and design. It uses the StripedHyena architecture and can model sequences at single nucleotide and byte-level resolution.

Evo is a biological foundation model capable of long-context modeling and design, employing a striped hyena architecture that models sequences at single-nucleotide byte-level resolution.

Chemical language foundation model proposed by IBM, supporting various tasks such as molecular representation conversion and quantum property prediction

TabPFN is a Transformer-based foundational model for tabular data. Through its prior data learning mechanism, it achieves outstanding performance on small-scale tabular datasets without requiring task-specific training.

A foundational model for tabular data, pretrained on synthetic datasets generated by mixing random classifiers

The Nucleotide Transformer is a set of foundational language models pre-trained on whole-genome DNA sequences, integrating genomic data from over 3,200 human genomes and 850 diverse species.

A multi-domain, multi-task language model designed to address a wide range of tasks in chemistry and natural language fields.

RNAErnie is a model for self-supervised pre-training based on non-coding RNA sequences. It uses a multi-stage masked language modeling objective to provide powerful feature representation capabilities for RNA research.

PlantCaduceus is a DNA language model pre-trained on 16 angiosperm genomes, utilizing Caduceus and Mamba architectures to learn evolutionary conservation and DNA sequence syntax through masked language modeling objectives.

A Transformer model pre-trained on large-scale single-cell transcriptome corpora for network biology prediction

A 500-million-parameter DNA sequence analysis model pre-trained on 3,202 genetically diverse human genomes

RNABERT is a pre-trained model based on non-coding RNA (ncRNA), employing Masked Language Modeling (MLM) and Structural Alignment Learning (SAL) objectives.

Caduceus-Ph is a DNA sequence modeling model based on the MambaDNA architecture, with a hidden dimension of 256 and a 16-layer structure.

AgroNT is a DNA language model trained on edible plant genomes, capable of learning universal representations of nucleotide sequences.

A 500M-parameter Transformer model pre-trained on the human reference genome, integrating DNA sequence information from over 3,200 diverse human genomes and 850 species

This is a bidirectional transformer model pre-trained on SMILES (Simplified Molecular Input Line Entry System) strings, primarily for molecular-related tasks.

A Transformer-based encoder-decoder model specifically designed for molecular representation using SELFIES

PlantCaduceus is a DNA language model pre-trained on the genomes of 16 angiosperm species, utilizing Caduceus and Mamba architectures to learn evolutionary conservation and DNA sequence syntax through masked language modeling objectives.

HyenaDNA is a long-range genomic foundation model pre-trained at single-nucleotide resolution with a context length of up to 1 million tokens.

GROVER is a pre-trained DNA language model specifically designed to understand and generate contextual representations of human genomic sequences.

A DNA sequence analysis model pre-trained on genomes from 850 species, supporting tasks such as molecular phenotype prediction

Caduceus-PS is a DNA sequence modeling model with reverse-complement equivariance, designed for processing long sequences.

Geneformer is a Transformer model pre-trained on large-scale single-cell transcriptome data, specifically designed for scenarios with scarce network biology data, enabling context-aware predictions.

HyenaDNA is a long-range genomic foundation model with a pre-training context length of up to 1 million tokens and single-nucleotide resolution.

ChemGPT is a generative molecular modeling Transformer model based on the GPT-Neo architecture, pretrained on the PubChem10M dataset.

BERT model trained on 50,000 SMILES strings for understanding and processing chemical molecular representations

This is a DQN agent model trained using stable-baselines3, specifically designed to solve reinforcement learning tasks in the MountainCar-v0 environment.

DNA sequence embedding model based on Transformer architecture, supporting sequence alignment and genomics applications

SegmentNT is a DNA segmentation model based on Nucleotide Transformer, capable of predicting the positions of multiple genomic elements in a sequence at single nucleotide resolution.

A self-supervised pre-trained foundation model for ECG analysis, supporting detection of 164 cardiovascular diseases

This model is a cheminformatics model pretrained on 10 million SMILES strings from the PubChem database, primarily used for molecular representation learning and chemical property prediction.

DrugGPT is a generative drug design model based on the GPT2 architecture, bringing innovation to drug design through natural language processing technology.

This is a reinforcement learning model based on the PPO algorithm, specifically designed to solve the balancing problem in the CartPole-v1 environment.

MOLT5-small is a molecular and natural language conversion model based on a pre - trained model, which can realize the mutual conversion between molecular structures and natural language descriptions.

ChemGPT is a generative molecular modeling tool based on the GPT-Neo model, specializing in molecular generation and research in the field of chemistry.

An autoregressive language model based on GPT2 architecture, specifically designed for generating drug-like molecules or embedding representations from SMILES strings

GENA-LM is an open-source foundational model family for long DNA sequences, based on a Transformer masked language model trained on human DNA sequences.

The PolyNC model achieves rapid and accurate prediction of polymer properties by integrating natural language and chemical language.

LeanDojo is a retrieval-augmented language model-based theorem proving system designed to enhance automated theorem proving by combining language models with retrieval techniques.

Uni-3DAR is an autoregressive model that unifies various 3D tasks, focusing on the generation and understanding of microscopic structures such as molecules, proteins, and crystals.