any solution for that? colab is blocking those commands and kaggle is auto turning off draft session in the middle.
The real fix isn’t “which cloud GPU” — it’s that Colab/Kaggle are AI/ML platforms, so policing what you run (hashcat included) is literally their job. Dedicated bare-metal hosts aren’t AI platforms — nobody’s watching your process list because that’s not their business.
The solid one — Hostkey:
→ Unmanaged dedicated GPU boxes (RTX 4090s, A100s), hourly or monthly
→ Actual hashcat forum regulars have pointed people here for years, specifically because nobody polices what runs on the box
→ It’s your box, not a shared marketplace host — no random eviction mid-crack
Job’s big enough that a dead session would actually hurt — Hashtopolis:
What it does
Splits one cracking job into chunks and spreads them across as many GPU agents as you rent. A box drops mid-run, that chunk just goes to another agent — no lost progress, no babysitting a session.
→ Official install guide for the server
→ Pre-built agent image (milz0/hashtopolis-hashcat-vast) if you want to feed it cheap marketplace GPUs
Don’t want to touch infrastructure at all:
Managed cracking services
OnlineHashCrack / HashCrack run the cluster themselves — send the hash, get results back. Both require you to own/be authorized for what you submit.
!