SCIP is a highly extensible solver and framework for combinatorial decision problems such as MIP (Mixed-Integer Programming), MINLP (Mixed-Integer Nonlinear Programming), and CIP (Constraint Integer Programs). It was created by Tobias Achterberg, and development traces back to the early 2000s (the first commit was on October 23, 2002, ~23 years ago).

It is modular and plugin-friendly, and the suite bundles several focused optimization flavors: SoPlex, Sequential object-oriented simPlex, is the linear programming solver backbone; ZIMPL is the compact modeling language that writes .lp/.mps files; UG provides the parallelization framework for branch-and-bound style solvers; GCG implements generic column generation and branch-price-and-cut; PaPILO supplies parallel presolve routines with multiprecision support. SoPlex moved to the Apache 2.0 license in late 2022, and ZIMPL is distributed under LGPLv3.

Releases up to and including version 8.0.2 are distributed under the ZIB Academic License, and from version 8.0.3, the full SCIP Optimization Suite is published under the Apache License 2.0. That license mapping makes it easier to use SCIP in permissively licensed projects from 8.0.3 onward.

According to Tobias Achterberg’s 20-year retrospective: textbook-simplex work, two-watched-literals propagation, generalized conflict analysis for infeasible LP relaxations, and experiments with history-based and ML-informed branching were important seeds that shaped reliability branching and many of SCIP’s heuristics and propagators.

SCIP also exposes bindings and solver interfaces such as PySCIPOpt for Python and NLPI interfaces to nonlinear solvers such as Ipopt, WORHP, and FilterSQP for MINLP work.