San Francisco AI evaluation startup Vals AI has closed a $40 million Series A led by Andreessen Horowitz at a $400 million post, cementing its role as the independent scorekeeper for frontier models as public benchmarks get saturated and leaked into training data.
Independent grades the labs actually cite
Founded in 2024 by Rayan Krishnan and Langston Nashold, Vals partners with reference institutions in law, banking, engineering and healthcare to build private, taxonomy-driven benchmarks that measure real professional work rather than trivia. Its scores are cited in model cards from OpenAI, Anthropic, Google, Meta and xAI, and its recent finance battery showed frontier models still fail 52% of real analyst tasks.
Round mixes flagship VCs with quant capital
The round was led by a16z with participation from existing investors 8VC, Pear VC and BloombergBeta, and new investors HRT Ventures and Next Ladder Ventures. Vals says revenue has grown 8x versus all of 2025, headcount has tripled in six months, and its customer base has doubled as enterprises scramble to quantify AI ROI.
New releases target coding, safety and government
Alongside the raise, Vals took Vals Smith generally available so any user can turn a GitHub repo into a custom coding benchmark, shipped a rebuilt Vals Index 2.0 with wider economic coverage, and rolled out frontier-risk benchmarks including a new cyber suite with academic collaborators, an RSI Index built with CoreWeave, and initial mental-health work. The company also flagged growing work with the U.S. Department of Commerce and Congress.
The Series A slots into a wider vote of confidence in the AI measurement layer, alongside Simile's $200M Series B and rising red-team spend such as Mindgard's $30M Series A.
Reporting based on coverage from Vals AI, TechCrunch, Crypto Briefing and Tech Funding News.