Every sword publishes the same attack. That is the finding.
Fextralife lists a melee attack figure for each of the eighteen swords it has a page for, and every single one of them is 51 or 45. Game8 lists twenty and every one of them is 27 or 21. Two sites, four numbers, twenty-two swords in the catalogue — the column a tier list would sort on has two values in it on either site, and neither site sorts by it. So a page that tells you the Kanzashi hits harder than the Bell Sword is telling you something no published source says. Below, the whole set is drawn on one scale so you can see the flatness rather than take our word for it; the second view re-encodes the same rows on the two fields that do vary.
22 swords · 2 distinct Fextralife values · 2 distinct Game8 values
Every bar is one sword’s published melee attack. Fextralife’s figures land on 51 or 45; Game8’s land on 27 or 21. Not one sword in this game is published as hitting harder than the rest of its own tier — so the number is not what you are choosing between.
Tap any sword for its passive, where it is, and how well sourced it is.
21 of these 22 rows are two wikis’ transcriptions and nothing more — no figure below was read off the game. The exception is Mechblade (Light), whose own panel reads 42 at +2, and it is the row that explains the rest: the two sites are quoting the same swords at different upgrade levels.
The two wikis are reading the same swords at different upgrade levels
The gap is not a transcription error, and it is not one site being wrong. LaserBolt’s early-power capture opens the sword list at 01:20 with an explicit upgrade suffix printed in every item’s own name — Mechblade (Standard)+2, Purifier Blade+2, Cirrus Sword+2 — and Mechblade (Light)+2 reads Melee Attack 42, which sits between Game8’s 27 and Fextralife’s 51. The passives climb too: that same +2 blade reads "increases the damage of attack 2 by 1.07× and critical chance by 20%" where both wikis publish 1.03× and no critical clause at all. So Fextralife is quoting enhanced values, Game8 is quoting base ones, the level is literally printed in the item name, and neither site says which one it read. Three points on the ladder — 27, 42, 51 — do not fix its shape, and nothing establishes that 51 is the top.
So pick on the element, which is the field that actually varies
Six damage types are spread across the sword roster — Slash, Rupture, Impact, Electric, Flame and Acid — and unlike the attack figure they are not interchangeable, because the elements page establishes that four of the eight types build a status and four do not. That makes "which sword is best" a question about what is in front of you rather than about the sword. The practical form of the question is: does this zone’s enemy set resist what I am carrying, and am I trying to stack a status or just to remove something. Three swords publish no damage type at all on either wiki, which is a gap in the sources rather than a claim that they are typeless.
What we are not telling you, and why
Nobody on this site has played far enough to rank these, and nothing published lets anyone else do it either. There is no damage-per-second measurement for any weapon in this game, no scaling table, no comparison of a fully-upgraded sword against another fully-upgraded sword, and no source that establishes what the enhancement ceiling even is. The roster itself is still contested: Gamer Guru counts 21 swords off the collection screen while completing Blade Collector, Game8 lists 20 and Fextralife 18, and Big Dipper appears on Game8 alone. Where a "best weapons" page elsewhere gives you a confident order, it is an order somebody assembled from the same two wikis you can read, plus an opinion — and this page would rather hand you the comparison and the caveat than the opinion.
The bows, crossbows and ammunition behave differently
Worth separating, because the two wikis mostly agree here and that agreement is itself evidence. Poison Arrows, Acid Bolts and Lightning Bolts are identical on both sites, and where the ammunition differs it is the stack limit and the buildup that move rather than a doubled attack value. The campfire’s Enhance Equipment menu, read off the client, lists three separate tracks — Sword Enhancement, Projectile Enhancement and Ranged Weapon Enhancement — so ammunition is Amber-enhanced like everything else. That makes the sword disagreement look like exactly what it is: one enhancement track that one wiki climbed and the other did not.