Core Keeper Codex
1.2.1.5 · verified 2026-09-07

Bosses

22 boss entities in Core Keeper 1.2.1.5, with health read from the game files and drop tables carrying the real selection weights.

BossLevelHealthDrop entries
Glurch the Abominous Mass22,45919
Ghorm the Devourer48,06121
Malugaz the Corrupted614,51327
King Slime624,188—
The Hive Mother624,67123
Haunted Hive Mother624,671—
Azeos the Sky Titan831,81318
Easter Bird Boss831,813—
Ivy the Poisonous Mass839,15518
Morpha the Aquatic Mass1084,55120
Omoroth the Sea Titan1190,77215
Scarab Boss Bomb Scarab122,535—
Igneous the Molten Mass13160,79916
Ra-Akar the Sand Titan13160,79916
Pyrdra the Fire Titan15136,99220
Druidra the Wild Titan15136,99219
Crydra the Ice Titan15136,99219
Queen of the Burrowed Sands15228,3208
Robot Boss16144,4149
Oblidra the Void Lord17160,0927
Core Commander17465,38513
Urschleim18249,82611

How Ghorm was identified

Ghorm is listed above under its real stats, and getting there took a detour worth stating. No entity in the files is named Ghorm, so this site previously said the encounter had no boss entity and guessed it was built from worm segments. That guess was wrong.

What settled it: every trophy is a boss-specific item, and across all 75 trophy objects that appear in any loot table, each is dropped by exactly one table — and wherever the boss is independently known, that table is its own. GhormTrophy appears in exactly one table, and it belongs to BossLarva. The same rule independently confirmed the Hive Mother's drop table, which had been an inference until now.

How to read a drop table — and why other sites' percentages are too low

Each boss table draws a fixed number of unique drops, weighted, without replacement. So an item's real chance is not its weight divided by the total: several draws happen, and each one is another shot at it.

The difference is not small. Azeos draws 6 unique items, so Scarlet Ore at weight 0.17 is 85.5% per kill, not the 21.5% that weight-over-total suggests — and the necklace is 29%, not 3.8%. Any site quoting the single-draw share is understating real drop rates roughly four-fold.

We publish the true figure. It is computed exactly rather than simulated: sampling without replacement is equivalent to an exponential race, which makes the probability a definite integral we can evaluate directly. The implementation is checked against brute-force enumeration of every possible draw sequence on every build, agreeing to about one part in a billion.