Models

AlphaFold 2

State-of-the-art protein structure predictor with benchmark-setting accuracy on monomers and broad adoption across biology.

Boltz-1

Biomolecular interaction modeling for protein complexes and small-molecule binding; joint structure and affinity inference.

Boltz-2

Successor to Boltz-1 with improved interaction modeling for complex structure and binding tasks; supports design workflows via scoring.

SimpleFold

First flow-matching based protein folding model using general purpose transformer blocks. Available in sizes from 100M to 3B parameters, challenges complex domain-specific architectures with competitive performance on standard benchmarks while enabling efficient deployment on consumer hardware.

ESM-1b

Evolution-scale protein language model family; enables fast structure prediction (ESMFold) and inverse folding for design (ESM-IF1).

EvoLLa + Llama3 8B

LLM adapted to protein property prediction with verbalized benchmarks, covering solubility, localization, PPI, and DTI tasks.

ProTrek

Tri-modal contrastive representations (sequence–structure–function) to navigate protein space and support design/annotation.

DeNovo-Pinal

De novo protein design guided by natural-language intent, generating candidates consistent with target specifications.

SaProt

Structure-aware protein language model with 3D-informed vocabulary that improves downstream property modeling and redesign.

SignalP 6.0

Signal peptide prediction across organism groups, including type classification and cleavage site identification.

AMP-Net

Deep learning framework for antimicrobial peptide discovery and prioritization to accelerate therapeutics design.

DeepLoc

Deep learning model for protein subcellular localization; widely used baseline (DeepLoc 2.0).

OpenMM

High-performance molecular dynamics engine for biomolecular simulation; GPU-accelerated and extensible.