Choose a backend

Solver backends are used by a prescriptive reasoner to solve a decision problem. This overview helps you choose the backend that best fits your problem type and scale.

Which backend should you use?

Use this table to choose the right backend for your problem:

Backend	Capabilities	When to use it
"highs"	LP / MILP Convex QP Continuous variables Discrete variables	Linear model or convex quadratic objectives. Bundled, no license required.
"gurobi"	LP / MILP QP / QCP NLP Continuous variables Discrete variables	Commercial solver for large/hard MIPs or you need QCP/NLP support. Requires license + backend enablement.
"ipopt"	Continuous NLP QP / QCP Continuous variables	Nonlinear continuous models.
"minizinc"	Constraint programming Discrete feasibility and optimization	CP-style constraints (for example, all_different) and discrete CSPs. Uses the Chuffed backend.

Note that:

If you are not sure, start with HiGHS. It is versatile and open source, so you can try it out without needing a license or setup.

Where to go next

Use HiGHS Bundled open-source default for LP/MILP and convex QP.

Use Gurobi Commercial solver for large/hard MIPs and broader model classes.

Use Ipopt Continuous nonlinear models (local optimality).

Use MiniZinc Constraint programming (CP) for discrete feasibility/optimization.