I notice that you have not published deployment fees on your web page. What do you expect those to look like?

On the Mittleman test problem set, it appears that CPLEX takes better advantage of parallelization, as Sebastian Pokutto noted. But on other test problem sets, the opposite may be true. We don’t have a direct comparison to CPLEX, but we have run 50 selected, hard-to-parallelize problems from the MIPLIB against Gurobi and XPRESS-MP with 1 core and 4 cores. On this test set, Gurobi’s outperformance vs. XPRESS-MP was significantly greater on 4 cores than it was on 1 core. The Gurobi Solver was engineered from the ground up for multi-core processors; while MIP Solver performance is always model-dependent, I’ll bet that further testing will show that Gurobi’s multi-core outperformance is the pattern rather than the exception.