Vals, backed by Andreessen Horowitz, is looking to become the gold standard for AI benchmarking

Vals AI was created to address perceived shortcomings in established AI benchmarking systems, which its founders say have lagged behind the rapid advancement of modern models. Rather than relying on publicly available tests that models can be tuned against, Vals keeps its specific test materials private and focuses on assessing a model's ability to carry out complex, domain-specific tasks. The company evaluates performance in sectors such as law, finance and coding and also runs specialized benchmarks in areas including recursive self-improvement, mental health, cybersecurity, biosecurity and the law of armed conflict.

Companies pay Vals to evaluate their models, a commercial model the founder likens to a student paying an organization to administer standardized exams. Vals argues that independent, confidential assessments can help developers identify weaknesses and track improvements over time, and that those results are increasingly used by buyers and decision-makers when selecting or acquiring models.

Since its founding, Vals has expanded rapidly. The startup reported that current revenue is eight times its level from the prior year and that headcount has grown from eight employees at the start of the year to about 25. The company recently raised $40 million in a Series A led by Andreessen Horowitz, following an earlier seed round led by 8VC and Bloomberg Beta. Vals plans further hires — an additional 10 to 15 people — and intends to move into a larger office.

Vals has also begun offering evaluations to federal agencies and is positioning its benchmarks as a potential standard for how AI capabilities are represented publicly. The company's founder said he expects rigorous evaluations to play a larger role as AI systems become more central to the economy and are disclosed in corporate filings or considered in investment decisions.