It is minimum requirements, and I usually try to use higher than minimum for safety margin.
That said I tend to consider load averages to only be actionable when I look into why they're getting where they are (I've seen load numbers get into the hundreds from uninterruptible sleep with an overstressed NFS server - which is a problem, but not the problem of the node that was seeing it).