Where is Vals AI headquartered?

Vals AI is headquartered in San Francisco, CA, USA. The company has a team of 11-50 people.

What does Vals AI do?

Vals AI operates in the AI Safety sector. Vals AI is an independent evaluator of artificial intelligence, building private, task-based benchmarks that measure whether AI models can do real work in domains like law, finance, coding, and frontier risk. Provides neutral, third-party AI model evaluation using privately held test sets to prevent benchmark leakage, measuring models on economically valuable real-world tasks rather than abstract exam-style knowledge.

When was Vals AI founded?

Vals AI was founded in 2024. It is based in San Francisco, CA, USA.

Who is the CEO of Vals AI?

Rayan Krishnan is the CEO of Vals AI. The company was founded in 2024.

Where is Vals AI located?

Vals AI is headquartered in San Francisco, CA, USA.

What funding stage is Vals AI at?

Vals AI is currently at the Series A stage.

How many employees does Vals AI have?

Vals AI has approximately 11-50 employees.

Vals AI

Category: AI Safety

Vals AI is an independent evaluator of artificial intelligence, building private, task-based benchmarks that measure whether AI models can do real work in domains like law, finance, coding, and frontier risk. Vals AI was founded in 2024. The company is led by Rayan Krishnan. Based in San Francisco, CA, USA. Team size: 11-50. Total funding raised: $45M. Latest round: Series A. Key investors include Andreessen Horowitz (a16z), 8VC, Pear VC, Bloomberg Beta, HRT Ventures (Hudson River Trading), Next Ladder Ventures.

Founded: 2024
Headquarters: San Francisco, CA, USA
Team size: 11-50
Total funding: $45M

Value proposition

Provides neutral, third-party AI model evaluation using privately held test sets to prevent benchmark leakage, measuring models on economically valuable real-world tasks rather than abstract exam-style knowledge.

Products and solutions

Vals Benchmarks (Finance Agent, coding, legal, cybersecurity, biosecurity, mental health, law of armed conflict), Vals Smith (custom coding benchmarks from any GitHub repo), Vals Index leaderboard, Frontier Risk Benchmarks (RSI Index with CoreWeave), Valkyrie (open-sourced distributed agentic benchmark runner), model library API, Public Benefits Bench (with Center for Civic Futures and Code for America), federal-agency evaluation program.

Unique value

Independent, leakage-resistant benchmarking using private test sets and domain-expert-designed tasks, positioning itself as the neutral measurement institution for the trillion-dollar AI market.

Target customer

Frontier AI labs (OpenAI, Anthropic, Google, Meta, xAI), enterprises selecting and validating AI models, and government/federal agencies.

Industries served

AI model evaluation/benchmarking, legal, finance, software engineering, cybersecurity, biosecurity, healthcare/mental health, government/public sector

Technology advantage

Proprietary private test sets (public validation set + private validation set + private test set) to prevent training-set contamination; open-sourced evaluation infrastructure (Valkyrie distributed agentic benchmark runner); automated scoring of generated work product; benchmarks cited in model cards by OpenAI, Anthropic, Google, Meta, and xAI.

How they differentiate

Uses private, non-public test sets to prevent benchmark gaming/leakage; evaluates real-world economically valuable tasks (law, finance, coding) rather than abstract knowledge; positions as a neutral third party independent of model builders.

Main competitors

Stanford HAI / academic benchmarks (HELM, SWE-bench), model-provider internal evals (OpenAI, Anthropic), other independent eval startups (e.g., Scale AI / SEAL, Patronus AI, LMArena)

Key partnerships

CoreWeave (RSI Index collaboration), Center for Civic Futures and Code for America (Public Benefits Bench), Stanford researchers, U.S. Department of Commerce and members of Congress (AI policy support), legal industry collaborations (law firms and legal research companies).

Notable customers

Frontier AI labs (OpenAI, Anthropic, Google, Meta, xAI cite Vals results in model cards), U.S. Department of Commerce, members of Congress, enterprise AI deployments

Major milestones

Founded 2024, $5M seed (July 2024) led by 8VC, benchmarks cited in model cards by OpenAI, Anthropic, Google, Meta, xAI, supported Department of Commerce and Congress on AI policy, launched federal-agency evaluation program, $40M Series A at $400M valuation led by a16z (Aug 2026), launched Vals Smith, Frontier Risk Benchmarks (RSI Index with CoreWeave), and Vals 2.0.

Growth metrics

Revenue grew 8x vs all of 2025; customer base doubled; team tripled from 8 to 25 in six months; $400M valuation at Series A.

Market positioning

Emerging category leader in independent AI benchmarking, aiming to become the "gold standard" measurement institution for AI, analogous to the College Board/SAT for models.

Geographic focus

United States (San Francisco), serving global frontier AI labs and U.S. federal agencies

About Rayan Krishnan

Ex-Software Engineer Intern at Palantir; worked at Microsoft and Stanford AI Lab as an undergraduate; Stanford University (AI master's). Co-founded Vals AI at age ~23.

Latest news about Vals AI

More AI Safety companies

Official website: https://www.vals.ai