OpenFold
OpenFold Consortium (Columbia, NVIDIA, SandboxAQ et al.)
Trainable, memory-efficient, GPU-friendly PyTorch reproduction of AlphaFold2. Includes full training code and data, enabling retraining and fine-tuning on custom datasets. Demonstrated AF2 reproducibility from scratch.
Best For
Retraining AF2 on proprietary data; understanding AF2 learning dynamics; commercial structure prediction
License
Open Source (Apache 2.0)
Strengths
- +Full training pipeline
- +Apache 2.0 license
- +Memory-efficient implementation
- +Consortium-backed
Limitations
- −Same accuracy ceiling as AF2
- −Significant compute for retraining
- −No co-folding (see OpenFold3)
R&D Pipeline Coverage
Related Tools
AlphaFold2
Google DeepMind
Predicts single-chain and multimer protein 3D structures from amino acid sequence using MSA-based deep learning. Set the modern benchmark on CASP14.
OpenFold3
OpenFold Consortium
Fully open-source AF3-architecture co-folding model. Full-stack release includes training data, weights, code, and evaluation scripts.
ColabFold
Steinegger Lab (Seoul National University)
Wraps AlphaFold2 with MMseqs2-based MSA generation, making AF2 runs 40-60x faster. Accessible via Google Colab or local install.
More in Structure Prediction
AlphaFold2
Google DeepMind
Predicts single-chain and multimer protein 3D structures from amino acid sequence using MSA-based deep learning. Set the modern benchmark on CASP14.
ColabFold
Steinegger Lab (Seoul National University)
Wraps AlphaFold2 with MMseqs2-based MSA generation, making AF2 runs 40-60x faster. Accessible via Google Colab or local install.
Boltz-1
MIT Jameel Clinic
First fully open-source model achieving AlphaFold3-level accuracy for joint structure prediction of proteins, nucleic acids, and small molecules.
Stay updated on OpenFold
Weekly newsletter covering AI tool releases, benchmarks, and what practitioners actually use.