Generative Models

The Generate section of the BioLM Console provides a catalog of Generative Models for designing new biological sequences and structures. These models include diffusion-based, inverse-folding, and causal language models that can be used to synthesize, optimize, or redesign proteins, antibodies, and nucleic acids.

Each model tile displays:

• Model Name – the identifier used throughout the platform and API.
• Description – an overview of what the model generates and how it can be applied.
• View API Docs – a link to the full API documentation, including example requests and responses.
• Cold Start Toggle – indicates whether a model instance is currently active or will need to be initialized before use.

Browsing Models

Generative models are displayed in a grid layout for quick access. Use the Tags dropdown to filter by model type, such as “protein generation,” “antibody design,” or “genome-scale DNA models.”

Example Model Types

Some representative models available in this section include:

• AntiFold – designs antibody sequences via inverse folding, optimizing CDR loops for affinity and structural integrity.
• Evo 2 1B Base / Evo 1.5 8k Base – autoregressive genomic language models capable of generating and analyzing DNA sequences at large scale, useful for genome design and CRISPR system analysis.
• ESM-IF1 – generates protein sequences from backbone coordinates, ideal for structure-based protein design.
• IgBert Paired – generates or scores paired antibody sequences using a model trained on heavy/light chain data.
• MPNN / Ligand ProteinMPNN – structure-conditioned sequence design models for proteins and protein–ligand complexes.

Using a Model

1. Select a model tile from the catalog.

   Click View API Docs to view usage details such as:

   • Input data types (e.g., PDB, FASTA, JSON)
   • Output schema and examples
   • Python SDK and REST API examples
2. Optionally toggle Cold Start if you’d like to pre-warm a model instance before making API calls.

Typical Use Cases

• De novo design of protein or antibody sequences
• Sequence optimization guided by structure or function
• Generation of novel DNA or RNA constructs
• Fine-tuning and design of biomolecules compatible with specific ligands, cofactors, or membranes

Notes

• Generative models can be used interactively via the console or programmatically via the BioLM SDK.
• Some models may require higher-tier plans or active compute balance for full access.