Sleep

The sleep spell of classic D&D was always a bit of a no-fun nuke spell that gave a magic-user one moment of glory in an adventure, but left them tossing daggers and holding the torch the rest of the session, or if they were lucky, ticking charges from a wand of something useful or casting from scrolls. Of course, the material costs of making scrolls meant no regular level one MU could have a bandolier of such scrolls handy to begin with, and even so, any sentient enemies would spot the pointy hat unrolling the parchment and make him their primary target...

Spell Card by Christopher Kosciuk

But I digress. What I am interested in doing is exploring ways to balance sleep. Some will argue there is no disparity, but granting NPCs the same cudgel and will to use it as the players, and a few TPKs later, if the game is still running, everyone will probably agree to rule-0 Sleep right out of existence.

First, simply allow saving throws for all targets. that will significantly cut into it's effectiveness and give both players and mooks a reasonable chance to remain upright. Apply a -1 to initiative checks for 2d4 rounds to simulate lassitude and away you go.

Further (or in lieu of allowing saving throws) allow the spell to affect 2HD of creature per level of the caster, and prevent success against foes whose HD exceed the casters level +1, or 5HD, whichever is less. I would treat half HD creatures as 0.5HD, not rounding, but otherwise disregard hp modifiers applied to HD. This allows 2 kobolds to be affected where one orc would otherwise be sent to dreamland. One might want to consider an upper limit on the number of HD affected, but leaving it open would provide for a significant non-lethal option that scales at the same rate as most other damage based spells (i.e. magic missile's notorious power scaling).

One may also consider introducing a variant spell to use either instead-of, or along-side the existing Sleep spell that instead of an area of effect, can be targeted on one or more individuals using either or both of the aforementioned mechanics. Targeted castings ability to take out the room sweeper and not the cannon-fodder should have restrictions. Clearly, this potential abuse of an already potent spell was appreciated when the spell was originally written, as it affects lowest to highest HD by default.

If one is using some sort of magic-point (mana) based casting solution, simply tie the number of targets affected to the number of magic-points expended on the spell. This allows the spell to remain a nuclear-option, while also allowing it to be used in more finessed situations for an appropriate cost (sleep only the guard on the left).