Largwit Evals
Private hacking tests for AI models.
Challenges no model has seen, so the score measures skill, not memory. Licensed per run or per year.
Public leaderboard
Coming soon.
A free public leaderboard of how frontier models score on a Largwit challenge set.
Work with Largwit
AI labs, data companies and AI security teams
Tell us what you're training, evaluating or shipping, and where it falls short.
Security experts and Gulf Arabic speakers
Tell us your field and experience. We're building a vetted bench of contractors.
We reply within two business days.