did the IBM online assessment + coding rounds a few weeks back while technically still employed, so i was doing this on lunch breaks. figured i'd actually write up what the format looks like since most posts i found were from 2022.
online assessment (OA): two coding problems on HackerRank. timed, 90 minutes total. difficulty was medium on a good faith scale, maybe medium-hard if you've been grinding easy-only. one was graph-adjacent (BFS/DFS would do it, nothing fancy), one was a string manipulation problem with some edge cases around unicode/special chars that i almost missed. no system design in the OA.
the OA also had a shorter 'cognitive + personality' section. multiple choice. i've never heard of anyone failing on that part so don't overthink it.
live coding round (if you pass OA): this was a 45-minute call, one interviewer, also HackerRank. problem was somewhere between LeetCode medium and easy-medium. interviewer was fine, interrupted a couple times to ask clarifying questions which felt more collaborative than hostile.
the thing nobody tells you: IBM does NOT demand optimal complexity proofs. i said 'this is O(n log n), i could get it to O(n) with a hash map but i want to get a working solution first' and the interviewer said 'totally fine.' they seemed way more interested in whether my code actually ran than whether i could whiteboard a proof.
verdict: easier than FAANG coding bars by a clear margin. if you can solve mediums consistently you'll be fine. don't over-prepare for the algo portion and end up under-prepared on behavioral, because IBM asks a LOT of behavioral questions in later rounds.