AI-Powered Cheminformatics

Derify develops state-of-the-art chemical language models for molecular property prediction, de novo drug generation, and efficient molecular information retrieval.

🤗Explore on Hugging FaceTry the API

Cheminformatics

Foundation Models

Information Retrieval

Our Models

State-of-the-art chemical language models designed for molecular information retrieval and property prediction.

🧪

ModChemBERT

FoundationModel

ModernBERT as a Chemical Language Model

View model card ↗

A ModernBERT-based chemical language model trained on SMILES strings using a multi-stage training pipeline. Achieves state-of-the-art performance on molecular property prediction benchmarks.

📐 768 hidden size

📏 256 max sequence length

💊 drug-like SMILES

ModernBERT
Molecular Property Prediction
Flash Attention
Apache 2.0

🔬

ChemMRL

EmbeddingModel

SMILES Matryoshka Representation Learning Embedding Transformer

View model card ↗

A state-of-the-art sentence transformer for generating molecular embeddings using Matryoshka representation learning. Enables flexible embedding dimensions for efficient similarity search and molecular information retrieval.

📐 1024 dimensions

📊 Tanimoto similarity

📏 512 max sequence length

Sentence Transformer
Embedding
Flash Attention
Apache 2.0

⚡

ChemRanker

RerankerModel

CrossEncoder for Molecular Reranking

A family of specialized CrossEncoder models for reranking a query SMILES against document SMILES. Complements ChemMRL for efficient similarity search and molecular information retrieval.

🎯 Reranking

📈 QED

📈 Similarity

🔍 IR pipeline

Cross Encoder
Reranker
Alpha
Flash Attention
Apache 2.0

Inference API

Empower your chemistry workflows with our high-performance Inference API. Generate precise molecular embeddings for similarity search, reranking molecules, and virtual screening using our state-of-the-art chemical language models.

  • Fast embedding generation
  • Batch processing support
  • Flexible quantization (int8, binary) for scalability

main.py

import requests

url = "https://api.derifyai.com/v1/models/Derify/ChemMRL/embed"
headers = {
    "X-API-Key": "YOUR_API_KEY",
    "Content-Type": "application/json"
}
data = {
    "smiles": ['CCO', 'CCN'],
    "normalize_embeddings": True,
    "precision": "float32"
}

response = requests.post(url, headers=headers, json=data)
result = response.json()
print(result)

Input

Endpoint

SMILES (one per line)

Options

Normalize Embeddings

Precision

Truncate Dimension

Leave empty to use full model dimensions.

API Response

No results to display. Press "Run" to fetch data.