Hard Evals
Research environments for frontier models

Hard Evals for investigative research.

100s of verifiable environments where today's frontier models struggle.

01

Investigative Journalism

can models reach the same conclusions as the best researchers today, given the same initial data?

02

Humanities research

can models demonstrate the judgement, intuition, and taste needed to uncover newsworthy findings across 1000s of sources?

03

Financial crimes

can models catch and track down the best-hidden financial fraud and money laundering using only publicly available data?

Contact us for access.

Email us