Vals raises $40 million to set new standard for AI model evaluation

Vals, a San Francisco startup founded in 2024, announced a $40 million Series A round led by Andreessen Horowitz. The funding follows a seed round last year that was led by 8VC and Bloomberg Beta. The capital will support the company’s effort to overhaul AI benchmarking, a practice that has become central to how firms validate model performance and market superiority.

Co-founder Rayan Krishnan, a 25-year-old former Palantir intern who studied at Stanford and worked at Microsoft and the university’s AI lab, says the company was created after he observed that traditional benchmarks were lagging behind rapid advances in model capabilities. “We were seeing a bunch of new, very capable models come to market quickly, and the academic benchmarks were not keeping up with that frontier advance,” Krishnan told TechCrunch. He argues that as AI integrates deeper into society, benchmarks must verify that models can deliver on the claims companies make.

Vals distinguishes itself by keeping its test materials private and by focusing on tasks that reflect real-world industry demands. Rather than generic knowledge quizzes, the startup evaluates models on complex functions in law, finance, software development and other sectors. It also probes potential negative outcomes, asking how models might behave if deployed without safeguards. The company’s roadmap includes benchmarks for recursive self-improvement, mental-health applications, cybersecurity, biosecurity and even the law of armed conflict, where it tests whether models can interpret the Geneva Convention.

Clients pay Vals to assess their models, a model Krishnan likens to students paying the College Board for the SAT. The service helps firms identify performance gaps and informs decisions about model acquisition. Vals’ revenue is now eight times what it was a year ago, reflecting growing demand for rigorous, industry-specific evaluation. The firm’s staff has expanded from eight employees at inception to 25, and Krishnan plans to add another 10 to 15 people while moving to a larger office space.

The company’s headquarters occupy two floors of a historic brick building on Folsom Street, once a brewery a century ago and now home to several tech startups. Vals recently launched a program that offers model evaluations to federal agencies, signaling an ambition to become a trusted benchmark provider for both private and public sectors.

Krishnan sees robust benchmarking as essential for the next wave of AI companies that are preparing to go public. “AI companies are starting to go public. SpaceX went public. Anthropic is slated for later this year. I suspect OpenAI will be public soon,” he said. “As AI models become a core part of the economy, the types of benchmarks and evaluations we do will drive their usage and become a central part of how these companies submit public filings or talk about prospective investments.”

The industry will be watching whether Vals can cement its methodology as the de-facto standard as AI continues its rapid expansion.