BORG3 is a submodule of the ARES3 framework. It comprises a set of complex non-linear models of large scale structure formation alongside with their adjoint gradient model. Their current list is:

- First Order Lagrangian Perturbation Theory (libLSS/physics/forwards/borg_lpt.hpp)
- Second Order Lagrangian Perturbation Theory (libLSS/physics/forwards/borg_2lpt.hpp)
- Full Particle-Mesh N-body simulation (libLSS/physics/forwards/borg_pm.hpp)

In addition to these forward models, BORG3 includes both specialized bias models and likelihoods and a generic bias framework. The generic framework supports the following data likelihood:

Additionally the following bias function are supported:

The specialized biases/likelihoods are the following (each model is detailed in the linked page):

- Poisson/Truncated power law
- Negative binomial/Truncated power law
- Gaussian/Linear bias Δ
- Generic bias framework
- Julia likelihood

Owing to the need of working in a non-shared memory environment (i.e. using MPI technology for parallelization), they have different communication technology developped to make the management of particles easy. We have notably developped a particle distribution mechanism which can be parametrized with any load balancing algorithm.