Vals Raises $40M to Fix AI Benchmarking

Vals, a San Francisco startup founded in 2024, has raised $40 million in a Series A led by Andreessen Horowitz to overhaul how AI models are evaluated. Unlike legacy benchmarks, Vals uses undisclosed, industry-specific tests to measure real-world performance in law, finance, coding, and emerging areas like cybersecurity and biosecurity. Companies pay Vals to test their models, similar to students paying for the SAT. The startup's revenue has grown eightfold year over year, its team has tripled to 25 employees, and it recently launched a program serving federal agencies. Co-founder Rayan Krishnan envisions benchmarking becoming central to AI public filings and investor disclosures.