Skip to navigation

Model Reference

The config each Model Hub model accepts when you submit a prediction batch.

When you create a prediction batch, each job in jobConfigs has a config. The shape of config depends on the model. This page lists the fields for each Benchling model.

Pass the model’s reference name (for example boltz2) as model.

Where config goes
{
"name": "My first batch",
"folderId": "lib_abc123",
"model": "boltz2",
"configSchemaVersion": 1,
"jobConfigs": [
{ "name": "Complex 1", "config": { "components": [ ... ] } }
]
}

Custom models your organization has installed aren’t listed here.

AlphaFlow-MD (alphaflow-md)

componentobjectRequired
The protein to run the model on.
useMsaboolean or nullOptional

This model always builds a multiple sequence alignment (MSA) for the input, so this must be true if set.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.

AlphaFlow-PDB (alphaflow-pdb)

componentobjectRequired
The protein to run the model on.
useMsaboolean or nullOptional

This model always builds a multiple sequence alignment (MSA) for the input, so this must be true if set.

numSamplesintegerOptional1-10Defaults to 1
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

AlphaFold2 (alphafold)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

This model always builds a multiple sequence alignment (MSA) for the input, so this must be true if set.

Boltz-1 (boltz1)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 1
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

Boltz-2 (boltz2)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

BoltzGen (boltzgen)

Send one flat config object. The protocol key chooses the design protocol, and the other keys depend on it. Both protocols design a binder against a single target, given in targets.

Protein binder

Designs a protein of a length you choose, set by minLength and maxLength. A protein target also accepts bindingResidues and notBindingResidues, as comma-separated residue numbers or ranges such as 5..7,13.

targetslist of objectsRequired
The target to design a binder for.
minLengthintegerRequired
Minimum length, in amino acids, of the designed protein.
maxLengthintegerRequired
Maximum length, in amino acids, of the designed protein. At most 1000.
protocol"protein-anything"OptionalDefaults to protein-anything

Design protocol. protein-anything designs a protein binder of a length you choose.

numDesignsintegerOptional1-1000Defaults to 8
Number of designs to generate.

Antibody binder

Designs an antibody onto predefined scaffolds. Set scaffolds to one or more scaffolds, each used at most once. This protocol doesn’t take minLength or maxLength.

targetslist of objectsRequired
The target to design a binder for.
scaffoldslist of stringsRequired
Antibody scaffolds to design onto. Each scaffold can be used once.
protocol"antibody-anything"OptionalDefaults to antibody-anything

Design protocol. antibody-anything designs an antibody binder onto predefined scaffolds.

numDesignsintegerOptional1-1000Defaults to 8
Number of designs to generate.

Chai-1 (chai1)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

DeepSP (deepsp)

antibodyobjectRequired
The antibody to run the model on.

DeepViscosity (deepviscosity)

antibodyobjectRequired
The antibody to run the model on.

ESMFlow-MD (esmflow-md)

componentobjectRequired
The protein to run the model on.
numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.

ESMFlow-PDB (esmflow-pdb)

componentobjectRequired
The protein to run the model on.
numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.

ESMFold2-Fast (esmfold2-fast)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
numSamplesintegerOptional1-10Defaults to 1
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

ESMFold2 (esmfold2)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 1
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

FlashABB (flashabb)

antibodyobjectRequired
The antibody to run the model on.

IntelliFold V2 (intellifold)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

Inductive: LogD (inductive:logd)

componentobjectRequired
The molecule to run the model on.

OpenDDE-AbAg (opendde-abag)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

OpenDDE (opendde)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

OpenFold (openfold2)

componentobjectRequired
The protein to run the model on.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

OpenFold3 (openfold3)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

Inductive: pKa (inductive:pka)

componentobjectRequired
The molecule to run the model on.

Protenix-v1 (protenix)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.

Protenix-v2 (protenix-v2)

componentslist of objectsRequired
The proteins and molecules that make up the complex to predict.
useMsaboolean or nullOptional

Whether to build a multiple sequence alignment (MSA) for the input. Omit to use the model's default.

numSamplesintegerOptional1-10Defaults to 5
Number of structures to sample.
seedintegerOptional0-4294967295
Random seed for reproducible results. Omit to use a random seed, which is saved with the config.