Vals Raises $40 Million to Redefine AI Benchmarking

Founded in 2024 by 25-year-old Rayan Krishnan, Vals targets the growing gap between rapid AI development and outdated evaluation systems. While many legacy benchmarks rely on public datasets that allow for model training contamination, Vals operates as a private arbiter. Companies pay the startup to evaluate their models on complex tasks in finance, law, and coding, effectively functioning like an SAT for artificial intelligence.

Krishnan, a former Palantir intern and Stanford researcher, argues that abstract testing fails to capture the true risks and capabilities of frontier models. His team is now stress-testing AI across high-stakes domains including cybersecurity, biosecurity, and the law of armed conflict. This approach has gained traction: revenue has grown eightfold over the past year, and the staff has expanded from eight to 25 employees. As major AI firms move toward public markets, Vals is positioning its evaluations as the necessary bridge between raw technical capability and public accountability.