Just thought I'd add the same benchmark on the latest Dyalog APL.
Dyalog APL/S-64 (17)
⍴+.×⌿?2 3000 3000⍴1e10
- size 3000: 6.6s, 0.5GiB RSS
- size 10000: 101s, 5.4GiB RSS
Conclusions:J's implementation is still the fastest at this particular task (matrix multiplication of huge matrices on a single CPU thread--granted, not the most significant of benchmarks.) Dyalog APL comes close behind. GNU APL and NumPy lag much more behind that.
GNU APL (1.7)
J (8.07) NumPy (1.13.3, blas/lapack 3.7.1) Conclusions:J's implementation is surprisingly performant! Easily beating NumPy on speed alone by a factor of 10 or more! (And I thought blas/lapack were already heavily optimized libraries!) J's memory usage is comparable to that of NumPy. GNU APL had the worst memory and cpu profile of them all.