Vals, a startup founded in 2024 to build more rigorous benchmarking systems for AI models, raised $40 million in a Series A round led by Andreessen Horowitz, following an earlier seed round led by 8VC and Bloomberg Beta. Co-founder Rayan Krishnan, a 25-year-old former Palantir intern who studied at Stanford and worked with Microsoft and the university’s AI lab, said the company emerged from observing that traditional academic benchmarks were failing to keep pace with rapidly advancing frontier models.
Unlike many benchmarking systems that rely on publicly available tests, Vals keeps its evaluation materials undisclosed to prevent companies from training models specifically to pass them. Rather than measuring general knowledge, Vals assesses models on their ability to complete real-world tasks across industries including law, finance and coding, while also evaluating potential negative outcomes in areas such as cybersecurity, biosecurity, mental health, and recursive self-improvement.
Krishnan said companies pay for evaluations similarly to how students pay to take standardized tests, using results to troubleshoot and improve their models over time. The company’s revenue has grown eightfold year-over-year, and its team has tripled to 25 employees, with further hiring and office expansion planned. Vals recently launched a program offering model evaluations to federal agencies, and Krishnan expects such benchmarks to play an increasingly central role as AI companies pursue public offerings.
