At least for strassen, the result is slightly less stable (increases the condition number by a constant factor). I think higham shows that any sub-cubic algorit

At least for strassen, the result is slightly less stable (increases the condition number by a constant factor). I think higham shows that any sub-cubic algorithm must have done conditioning issues, but that in practice they're not that bad for most matrices.