Best Sword Enchantments in Minecraft

On Java Edition a sword can carry seven worthwhile enchantments at the same time, and only one of them is actually a decision. Sharpness V, Looting III, Unbreaking III, Mending, Sweeping Edge III and Fire Aspect II all sit on the same blade without complaint, and Knockback II fits as well — though most players leave that one off deliberately. The single contested slot is the damage one, where Sharpness, Smite and Bane of Arthropods lock each other out and the game gives you exactly one.

The short answer is Sharpness V for anything you will realistically fight. The exception is bigger than it looks: against undead mobs, Smite V adds more than four times the damage Sharpness V does. If a sword's job is a zombie or skeleton farm, Smite is not a novelty pick, it is the correct one.

The numbers below are the current ones rather than the ones still circulating in old videos, and Java and Bedrock genuinely differ. Bedrock uses a different Sharpness formula, rounds Smite down, and does not have Sweeping Edge at all.

Version checked

Verified against Minecraft: Java Edition 26.2 ("Chaos Cubed", released 16 June 2026) and Bedrock Edition 26.45 (28 August 2026), the current releases as of 12 September 2026. Every level cap, damage value and conflict rule here was read from minecraft.wiki on that date rather than from memory.

The set worth building

That is the complete list — ten enchantments, of which eight can legally coexist. Our enchantment calculator holds every level to its real cap and greys out a row the moment something you have already picked conflicts with it, then hands you the matching /give or /enchant command, which is handy for testing a loadout before you spend the levels earning it.

The only real decision

Sharpness, Smite and Bane of Arthropods are mutually exclusive on a sword. Here is what that choice is worth on a diamond sword, which deals 7 damage bare:

Sharpness adds a flat +3 against everything. Smite adds +12.5, but only against zombies and their variants, skeletons and their variants, phantoms and withers. Bane of Arthropods adds the same +12.5 against spiders, cave spiders, silverfish, endermites and bees, and also inflicts Slowness IV for a random 1–1.5 seconds at level I, rising to a maximum of 3.5 seconds at level V.

For general play Sharpness wins because it applies to creepers, players, the Warden and everything else Smite ignores. For a dedicated farm, Smite V takes a diamond sword from 7 damage to 19.5 — almost double what Sharpness V manages. Bane of Arthropods is the weakest in practice: the arthropod list is short and mostly harmless.

The enchanting table is biased towards Sharpness independently of that: Sharpness carries a selection weight of 10 against 5 for Smite and 5 for Bane of Arthropods, so it comes up roughly twice as often.

An anvil will take a conflicting book and bin it

Since Java Edition 1.12.2 the anvil no longer refuses a conflicting combination. Put a Smite book on a Sharpness sword and the operation succeeds — the wiki notes that "the sacrifice item's enchantment is lost", and Java additionally adds one level to the cost "for every incompatible enchantment on the target". You pay extra to throw the book away. The /enchant command is stricter and refuses outright, because it is "subject to the same restrictions as an anvil" and validates before applying.

Looting, Fire Aspect, and the trap between them

Looting raises the maximum of most common drops by one per level and increases the chance of rare drops and equipment drops by one percentage point per level. The percentage point matters more than it sounds: a geared mob has an 8.5% chance on Java (25% on Bedrock) of dropping what it is wearing, plus 1 point per Looting level, so Looting III takes that to 11.5%.

Two mechanics are worth knowing. First, Looting applies to any kill made while the enchanted sword is in your main hand, not only to kills made with the sword — so firing a bow and switching to the sword before the arrow lands still gets the bonus. Second, and this is the trap: mobs killed by the burn from Fire Aspect do not get Looting applied to their drops. If you are farming, Fire Aspect actively costs you loot.

Looting also does nothing for sheep wool, evoker totems, wither skeleton skull counts (though it does raise the chance of one dropping at all), warden sculk catalysts, or kills made by a tamed wolf.

Unbreaking and Mending, together

Unbreaking gives a 100% ÷ (level + 1) chance that a use costs durability, which at level III means three uses in four are free. A diamond sword's 1,562 uses become an effective 6,248; netherite's 2,032 become 8,128.

Mending then repairs 2 durability per point of experience in each orb you collect, and it takes priority over your experience bar — if anything equipped is damaged, the orb goes there first. It picks one damaged item at random per orb regardless of how badly damaged each is, so a full Mending kit repairs slowly and unevenly. With both on the same sword, the sword stops being consumable.

Sweeping Edge is Java-only

Sweep damage is 1 + Attack_Damage × (Level / (Level + 1)). An iron sword with only Sweeping Edge I deals 4 to each surrounding mob; a diamond sword with Sharpness V and Sweeping Edge III deals 8.5. Without it, a sweep deals 1. Sweeps fire when you are grounded, not performing a sprint-knockback attack, with the attack charge at 84.8% or above, and they hit entities within a 1 × 0.25 × 1 block box around your target.

Bedrock has no Sweeping Edge, and no sweep attack to enchant. It also uses 1.25 × level rounded down for Sharpness, giving +1/+2/+3/+5/+6 instead of Java's +1/+1.5/+2/+2.5/+3, and rounds Smite and Bane of Arthropods down so levels I, III and V are slightly weaker than Java's. Bedrock sword base damage is one higher across every tier. If your group is mixed, see Is Minecraft Crossplay.

Getting there without hitting "Too Expensive!"

Fifteen bookshelves put the bottom enchanting-table slot at exactly 30 levels, which is the ceiling. Even so, a diamond sword cannot roll Sharpness V there — the wiki is explicit that "the wooden, stone, copper, iron, diamond, and netherite tiers cannot receive Sharpness V through an enchanting table" and that "the golden tier is the only tier that is able to receive Sharpness V directly through an enchanting table". Gold's enchantability is 22 against diamond's 10, and that is the whole reason. So Sharpness V on a real sword comes from a book and an anvil.

Also worth knowing: the table's three offers come from an enchanting seed stored as XpSeed in your own player data, and it only rerolls when you actually enchant something. Walking to a different table with the same number of bookshelves shows you the same offers.

The one genuinely server-side thing here

Nobody needs a server to enchant a sword — this is all single-player mechanics. But XpSeed lives in each player's own save data, so on a managed Minecraft server two players at the same table see different offers, and one player enchanting does not reroll anyone else. And if you want to change the caps rather than work within them, Java stores every enchantment as JSON at data/<namespace>/enchantment in a data pack — max_level, weight and exclusive_set are all editable there, and that is the only supported way to do it.

Applying six books one at a time is the expensive route, because each anvil use roughly doubles the item's prior work penalty. The penalty is 2n + 1 where n is the larger of the two items' existing penalties, which walks an untouched sword up 0, 1, 3, 7, 15, 31 (Java stores it as the item's repair_cost):

That is 83 levels, and step six costs 33 of the 39 an anvil will accept before it refuses with "Too Expensive!". One more enchantment in that order and the anvil refuses outright. The fix is the rule the wiki states directly: "prior work penalties are minimized by combining two items with equal penalties if possible" — merge books into books first, then apply one combined book to the sword.

If a sword is already too expensive, a grindstone is the escape hatch. It strips every non-curse enchantment and refunds a random 50% to 100% of the combined minimum enchantment levels as experience — Sharpness V averages 35 back — and crucially the output has no prior work penalty at all. You get a clean sword and can start again.

What to Read Next

• Minecraft Server Commands: The Complete Op & Admin List — where /enchant and /give sit, and the op levels they need
• Datapacks and Resource Packs — how to load the JSON that defines an enchantment
• Getting started with your Minecraft server — first boot, if you are building a world to test in
• Curse of Vanishing — the one sword enchantment you never want, and why you cannot remove it
• Best Level for Diamonds — you need the sword before you need the enchantments
• Trial Chambers — one of the better sources of enchanted loot

---

Made with 💜 by GameServerKings