Powerful and flexible
PoET-2’s architecture generalizes to families and properties it has never seen.
Protein language models capture evolutionary patterns shaping protein structure and function
OpenProtein.AI — The AI Cloud for Protein Design
Developed by the pioneers of protein language models
Sign UpThe AI Cloud for Protein Design
Generate diverse, functional sequences de novo given sequence and structural constraints
Learn more →
From evolution to design: a complete protein engineering workflow
Learning from billions of years of evolution
...to drive end-to-end protein engineering
Design protein sequences de novo
Analyze the fitness landscape and prioritize variants
Zero shot variant effect prediction
Evaluate the functional impact of novel variants in silico - including indels
Learn from your mutagenesis data
Train specialized models on your own data
Screen libraries with specialized models
Rank and filter sequences to focus on the most promising candidates
Design combinatorial variant libraries
Optimize variants for multiple properties
Generate sequences
Variant effect prediction
Train models
Predict variants
Design libraries
How OpenProtein.AI helps you
From de novo design to optimization, OpenProtein.AI provides the tools and AI infrastructure for protein design. Accelerate your design-build-test cycles and maximize the value of every data point.
Reduce costs more than 60%
Accelerate your projects more than 3x
Eliminate guesswork
Built for any protein, any property
Type
- Antibodies
- Nanobodies
- Enzymes
- Structural proteins
- Receptors
- Peptides
Properties
- Activity
- Stability
- Expression
- Humanization
- Manufacturability
- Immunogenicity
End-to-end protein design and optimization tools for biologists and bioinformaticians
User-friendly web interface, no coding required
End-to-end protein engineering suite, easily accessible through a powerful graphic interface.
Scalable APIs for high-throughput protein engineering
Programmatic access to models for design, optimization, and analysis. Built to scale.
AI infrastructure for the new world of protein engineering
Integrate models, data systems, and ML frameworks across experimental and computational teams
Directly leverage ML advances
Accelerate research with self-service, no-code access to your team’s ML models, instantly integrated directly into existing workflows
Deploy custom models directly to biology teams
Build better models faster with managed infrastructure, and seamless deployment to biology teams
Enabled by our proprietary foundation models
and open source model library
State-of-the-art model for protein design
PoET-2
Pushing the boundaries of protein language models
Leading zero-shot performance
State-of-the-art zero-shot performance across benchmarks spanning fitness, stability and binding. Design protein variants with no assay data required.
Predict protein properties
Train more accurate sequence-property predictors with 10x less data, cutting assay costs and accelerating iteration cycles.
A complete library of the best protein AI models
Access protein foundation and generative models all in one place. Use these models to design new proteins, fine-tune them with new data, or run in silico screening with our APIs or GUI. Compare between models to find the best model for your application.
PoET-1 and PoET-2
Embed, score, and generate sequences conditioned on a protein family. PoET-2 also accepts sequence and structural inputs as constraints and context.
ESM-1 and ESM-2
Encoder-only transformer models trained on large protein sequence datasets. Supports embedding, scoring, and attention maps.
Citation: Lin, et al., Evolutionary-scale prediction of atomic-level protein structure with a language model. Science 379, 1123-1130 (2023). doi:10.1126/science.ade2574
ProtT5
Encoder-decoder transformer models trained on large protein sequence datasets. Primarily used for embedding and representation learning.
RFdiffusion
Diffusion model for novel backbone and complex structure design. Supports motif scaffolding and binder design.
Citation: Watson, et al., De novo design of protein structure and function with RFdiffusion. Nature 620, 1089–1100 (2023). doi:10.1038/s41586-023-06415-8
BoltzGen
Generates 3D structures and conformational ensembles conditioned on sequence, motifs, or interaction constraints.
Citation: Stark, et al. Toward universal binder design with BoltzGen. bioRxiv 2025.11.20.689494 (2025). doi:10.1101/2025.11.20.689494
Boltz-1 and Boltz-2
Predict static and dynamic biomolecular structures and binding affinity to support structure-based drug discovery.
Citation: Passaro, et al. Boltz-2: towards accurate and efficient binding affinity prediction. bioRxiv 2025.06.14.659707 (2025). doi:10.1101/2025.06.14.659707
AlphaFold2
Predicts 3D protein structures from amino acid sequences using deep learning and evolutionary information.
Citation: Jumper, et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021). doi:10.1038/s41586-021-03819-2
ESMFold
Predicts 3D protein structures directly from sequences using transformer-based language models trained on large-scale datasets.
Citation: Lin, et al. Language models of protein sequences at the scale of evolution enable accurate structure prediction. Science (2023). doi:10.1126/science.ade2574
Protenix
Lightweight transformer for fast embedding, scoring, and representation learning on downstream ML tasks.
Citation: Chen, et al. Protenix: advancing structure prediction through a comprehensive AlphaFold3 reproduction. bioRxiv 2025.01.08.631967 (2025). doi:10.1101/2025.01.08.631967
Minifold
Compact AlphaFold-style architecture for rapid structure predictions with reduced compute on small–medium proteins.
Citation: Alcaide, E. (2019). MiniFold: a DeepLearning-based Mini Protein Folding Engine. GitHub. doi:10.5281/zenodo.3774491
RosettaFold3
Integrates sequence, structure, and diffusion-based reasoning for improved complex prediction and conditional design.
Citation: Baek, et al. Accurate prediction of protein structures and interactions using a three-track neural network. Science 373, 871–876 (2021). doi:10.1126/science.abj8754
Check it out →
...and more
Backed by publications and real world case studies
Understanding protein function with a multimodal retrieval-augmented foundation model
Timothy F. Truong Jr, Tristan Bepler
NeurIPS 2025
Machine learning optimization of candidate antibody yields highly diverse sub-nanomolar affinity antibody libraries
Lin Li, et al.
Nature Communications 2023
PoET: A generative model of protein families as sequences-of-sequences
Timothy F. Truong Jr, Tristan Bepler
NeurIPS 2023
Maximize the value of your data and build predictions tools tailored to your proteins and projects
Combine proprietary data with foundation models to train specialized models tailored to your needs. Deploy these models to accelerate new projects.
Sign up now →
Meet the demands of every protein engineering project
No usage barriers - transparent
pricing
Work with the top experts in protein ML
Plug and play system - no setup required
Best practices workflows out of the box
Connects experimental and computational teams
Data protection and IP ownership
Your data stays private, encrypted, and siloed
We safeguard your data with encryption, full account isolation, and secure cloud infrastructure. No one accesses your data but you.
Exclusive ownership of your outputs
Retain full IP rights to your dataset, designs and output. No royalties, no licensing fees.
Get started with OpenProtein
