a16z leads 400M valuation to test AI on real-world tasks
August 14, 2026 — Tech Funding News
Vals AI closed a 400 million valuation, led by Andreessen Horowitz. Existing backers 8VC, Pear VC, and Bloomberg Beta returned; new investors HRT Ventures and Next Ladder Ventures.
Problem
Academic benchmarks are saturated, leak into training data, or become optimization targets. Models ace leaderboards but fail on messy real-world multi-step agent work.
Solution
Vals pairs domain experts (law, finance, healthcare, coding) with automated grading against professional standards. Test sets stay private; limited runs prevent contamination. Can benchmark new models within hours; retires tests when they stop separating models (e.g., swapped CorpFin for Excel-modeling test in May).
Evaluations cited in model cards from OpenAI, Anthropic, Google, Meta, xAI. Enterprises use scores for production model selection.
Founders Rayan Krishnan (CEO) and Langston Nashold (CTO) met at Stanford. Team alumni from Palantir, Microsoft, NVIDIA, Meta, Hudson River Trading.
New products with funding
- Vals Smith — custom coding benchmarks from customer GitHub repos
- Frontier-risk benchmarks (cybersecurity, mental health, AI safety)
- Vals Index 2.0 — broader economy measurement
Revenue up 8× in 2025; customer base doubled; team tripled in six months. Prior $5M seed from 8VC, Bloomberg Beta, Pear VC.
Jennifer Li (a16z GP) compared Vals to Moody’s for credit markets — independent scorekeeper when sellers know more than buyers.