# Vals AI

> Vals AI is a San Francisco company that independently benchmarks AI models, agents and applications on professional tasks in finance, law, healthcare, coding and other fields. It publishes comparisons and sells evaluation infrastructure, including custom coding benchmarks built from a customer’s own repository.

- **Founded:** 2024
- **Headquarters:** San Francisco, United States
- **Founders:** Rayan Krishnan (Co-founder and CEO), Langston Nashold (Co-founder and CTO)
- **Products:** Vals Index and benchmark leaderboards, Vals Smith, Domain benchmarks, Valkyrie, Enterprise and public-sector evaluations
- **Notable:** Published its first public benchmark in 2024 covering 15 language models across four domains., Its benchmark results have been cited in model cards from OpenAI, Anthropic, Google, Meta and xAI, according to Vals., Its February 2026 healthcare study found leading models scored about 88% on clinical documentation and about 56% on diagnosis coding.

## Products & services

- **Vals Index and benchmark leaderboards** — Public cross-model comparisons covering task performance, cost and latency across professional and technical domains.
- **Vals Smith** — Creates custom coding benchmarks from public or private GitHub repositories by converting merged pull requests into reproducible tasks with hidden tests.
- **Domain benchmarks** — Private and public evaluations including Finance Agent, Legal Research Bench, MedScribe, MedCode and frontier-risk tests.
- **Valkyrie** — Open-source orchestration infrastructure for scalable, reproducible agent evaluations.
- **Enterprise and public-sector evaluations** — Custom model, agent and application testing for enterprise teams and government users.

## Achievements

- Published its first public benchmark in 2024 covering 15 language models across four domains.
- Its benchmark results have been cited in model cards from OpenAI, Anthropic, Google, Meta and xAI, according to Vals.
- Its February 2026 healthcare study found leading models scored about 88% on clinical documentation and about 56% on diagnosis coding.
- Announced a $40 million Series A at a $400 million valuation in August 2026.
- Released the open-source Valkyrie evaluation framework.

## Latest updates

- **2026-09** — Published an audit of benchmark cheating across BioMysteryBench, Terminal-Bench 2.1 and SWE-bench Verified.
- **2026-09** — Published research on AI safety in teen conversations and model evaluations across its suite.
- **2026-08** — Announced $40 million Series A, Vals Smith general availability, frontier-risk benchmarks and Vals Index 2.0.
- **2026-02** — Released MedScribe and MedCode healthcare benchmark findings.

## Links

- Website: https://www.vals.ai
- LinkedIn: https://www.linkedin.com/company/vals-ai/
- Twitter/X: https://x.com/ValsAI
- GitHub: https://github.com/vals-ai

---

Profile page: https://yespress.io/vals-ai
Published by YesPress — https://yespress.io
Last updated: 2026-09-30