[leethargo]

To add to the above: The high-performance commercial solvers typically offer a free (as in beer) licence for academics (students and researchers), so this subgroup has a smaller incentive to develop a competing solver.

Similarly, researches who do spend their time implementing solver algorithms and running tedious computational experiements (the work that the software vendors put in) have historically had difficulties getting academic credit for their work, because the journals favored theoretical work.

That being said, with HiGHS and SCIP, we have two open-source solvers developed in an academic setting, with a lot their graduates joining commercial software vendors. So it's not like these are two completely separate worlds.

[npalli]

Sadly though, the Open source solvers seem to perform pretty badly compared to commercial solvers.

https://plato.asu.edu/bench.html

[leethargo]

True, especially on difficult problems. Open-source can be good enough for many practical applications, though.

In my opinion, the gap in performance is less important, but the commercial offerings are typically more robust/reliable.