Benchmarking is the industry's gold standard for validating AI models, but legacy systems are often outmatched by the pace of innovation. Enter Vals, a startup co-founded by 25-year-old Rayan Krishnan, which aims to revolutionise the way we measure AI performance.

Unlike traditional benchmarks that measure general knowledge, Vals evaluates models on complex industry-specific tasks, such as law and finance, to ensure they can perform real-world work as well as humans. Krishnan explains, 'We’re looking at the real impacts of models, not just positive outcomes.'

With backing from major investors like Andreessen Horowitz, Vals has raised $40 million and now boasts a team of 25. The startup’s unique approach is gaining traction, with companies paying to test their models and federal agencies seeking model evaluations. Vals sees its system as the future of AI benchmarking, paving the way for public trust and transparency as AI becomes integral to the economy.