Vals Raises $40 Million to Standardize AI Evaluation

San Francisco-based startup Vals has secured a $40 million Series A funding round led by Andreessen Horowitz to expand its proprietary AI evaluation platform. Founded in 2024 by Rayan Krishnan, the company previously raised a seed round backed by 8VC and Bloomberg Beta. The fresh capital arrives amid rapid growth for the startup, which has seen its revenue surge to eight times its previous year's total. To support this expansion, Vals has tripled its workforce from eight to 25 employees this year, with plans to hire an additional 10 to 15 people and relocate to a larger office.

The startup aims to address a critical flaw in traditional AI benchmarking: public tests that allow developers to inadvertently or intentionally train their models on the exam questions. To prevent this, Vals keeps its evaluation materials strictly confidential. Rather than testing abstract intelligence or general knowledge, the platform measures how models perform on complex, domain-specific tasks. Its evaluations span practical industries like law, finance, and coding, alongside specialized fields such as mental health, cybersecurity, biosecurity, recursive self-improvement, and the application of the Geneva Convention in armed conflict.

For AI practitioners and enterprise buyers, this shift to closed, rigorous testing provides a more reliable metric for purchasing and deployment decisions. Companies pay Vals to audit their models, using the feedback to troubleshoot and refine performance. The startup has also launched a new initiative to provide model evaluations for federal agencies. As AI companies prepare for potential public listings, Krishnan believes these independent audits will become essential for establishing public trust and satisfying regulatory requirements.