Vals independently tests frontier AI models on real-world tasks, building private, domain-specific benchmarks that resist leakage, cherry-picking, and academic proxy metrics.
Products & media
Vals AI Legal Research Bench
An independently run collection of live leaderboards measuring leading AI models across economically valuable tasks, frontier risks, cost, and latency.
An independent government evaluation service that builds expert-backed benchmarks for public benefits, cybersecurity, policy, and national-security AI use cases.
A tool that converts public or private GitHub repositories into validated custom benchmarks for comparing coding models and agents.
An independent evaluation library reporting task-level performance of commercial AI applications, including legal research systems and other vertical products.
Cosigns
“Vals takes a fundamentally different approach to evaluation: test models on the work people actually want them to do, not on contrived exams.”
Team
Founders
2 people
Jobs
See all jobs for Vals here.
Funding
All fundraising announcements are based on public sources. Please contact Cosign if you believe there are any discrepancies, or claim the profile to edit.
Data current as of August 13, 2026
Leads
Participating Firms
Angels