Vals has raised $40 million in Series A funding led by Andreessen Horowitz as it builds independent benchmarks designed to measure how AI models perform on real professional work rather than traditional academic tests. Founded in 2024, the startup argues that many widely used AI benchmarks are becoming less useful as models improve and companies learn to optimize specifically for published tests.
The funding follows an earlier seed round led by 8VC and Bloomberg Beta and comes after a period of rapid growth. Vals says its revenue is now eight times higher than a year ago, while its team has expanded from eight employees at the beginning of 2026 to 25.
Testing Models on Real Professional Work
Traditional benchmarks often evaluate whether a model can answer standardized questions or pass exams, but Vals co-founder Rayan Krishnan argues those results do not necessarily show whether the same model can perform useful work. Vals instead creates evaluations around practical tasks in fields including law, finance, coding, healthcare and mathematics.
Its benchmark methodology also keeps the final test datasets private, reducing the risk that evaluation questions appear in model training data or that developers deliberately optimize models against known tests. Domain specialists help define tasks and grading standards, allowing Vals to compare AI outputs with the quality expected from human professionals.
The company has expanded into higher-risk areas as model capabilities increase. Krishnan said Vals is developing evaluations covering recursive self-improvement, mental health, cybersecurity, biosecurity and the law of armed conflict, including whether models can correctly apply provisions of the Geneva Conventions.
Companies pay Vals to evaluate their systems, giving developers an independent measurement they can use to identify weaknesses and compare models. Those results can also help businesses deciding which models to deploy for specific workflows rather than relying entirely on benchmark scores published by the model developers themselves.
AI Evaluation Becomes a Larger Business
Andreessen Horowitz, which led the Series A, said Vals focuses on evaluating whether models can complete the work people actually need them to perform, including legal research, financial analysis and software development. The approach comes as AI systems move beyond answering questions toward agents capable of completing longer, multistep tasks.
Vals is also expanding beyond commercial customers and recently launched a program providing model evaluations to U.S. federal agencies. Krishnan expects independent testing to become increasingly important as AI systems spread through businesses, government agencies and other institutions.
The company plans to hire another 10 to 15 employees and move into a larger San Francisco office. Krishnan said he expects standardized independent evaluations eventually to influence how AI companies communicate performance to customers, investors and public markets as the industry matures.
Featured image credits: Vals AI
For more stories like it, click the +Follow button at the top of this page to follow us.