Thursday, January 29, 2009

Gurobi vs Cplex benchmarks

> What are your first impressions of Gurobi? How does it compare to CPLEX for MIP? Have you done any benchmark analysis?

Here are some benchmarks:

However I would always urge you to try out solvers on your particular models as they may behave very differently than the benchmark models. Drop me a note if you need me to run some models or need access to evaluation systems.


  1. We have the Gurobi Solver available for download right now (15 day free trial with no other restrictions) at You can run the Mittleman benchmarks yourself, or solve your own models in MPS, LP and OSIL format using a little program GurobiEval that we include. You can create models from scratch, and solve them with Gurobi, either in Excel with our modeling system, Risk Solver Platform, or in C, C++, C#, VB6, VB.NET, Java or MATLAB using our Solver Platform SDK.

    It's very true that LP/MIP Solver performance depends considerably on the model or set of models tested. I understand the Gurobi Solver was designed from the ground up for multi-core processors, but its relative performance with 4-cores vs. 1-core on the Mittleman problems doesn't really reflect this. On a selection of 50 MIPLIB problems we chose that seem hard to parallelize, the Gurobi Solver's relative performance with 4-cores vs. 1-core was much better. It really pays to try these Solvers on your own model (and work to improve your own model formulation).

  2. No doubt Gurobi will be a valuable addition to the available commercial LP/MIP solvers. Of course building a solver from scratch has its advantages (after having learned from previous experiences).

  3. Gurobi is now also available as solver under Microsoft Solver Foundation (it is the default MIP solver; see as solver under GAMS (beta 23.0).

  4. You may want to check this post for a simple IDE to gurobi, and this one if you are interested in combining Gurobi with iPython.