vals.ai

Vals AI operates as an independent, third-party evaluation and benchmarking platform for large language models (LLMs) and artificial intelligence applications. The company provides detailed reporting on how AI models perform on real-world, industry-specific tasks, addressing a gap in standardized, unbiased performance metrics. Vals AI develops and hosts leaderboards and benchmarks for specialized domains including finance, law, healthcare, and software development. These benchmarks are created in collaboration with domain experts and utilize private datasets to prevent test leakage and ensure the integrity of the results. The platform assesses models on various capabilities such as tool use, multi-modality, long-context reasoning, accuracy, latency, and cost.

The company was founded in San Francisco in 2023 by Rayan Krishnan (CEO) and Langston Nashold (CTO). The two co-founders studied computer science together at Stanford University, where they had already begun collaborating on real-world measurement problems before establishing Vals AI. The founding team's background includes experience at prominent tech companies and research institutions like NVIDIA, Meta, Microsoft, Palantir, and Hudson River Trading. Initially bootstrapped, Vals AI reported an estimated annual revenue of $1.3 million as of August 2026. In August 2026, the company announced a $40 million Series A funding round led by Andreessen Horowitz, with participation from existing investors 8VC, Pear VC, and Bloomberg Beta, reaching a valuation of $400 million.

Vals AI serves AI research labs and enterprise engineering, product, and leadership teams that require rigorous, neutral information to make high-stakes decisions about AI model adoption and deployment. The business offers its evaluation infrastructure to private labs and enterprise clients, allowing them to assess AI applications on their own proprietary data and tasks. The company also produces public benchmarks, some of which are open-sourced, such as the 'Finance Agent' benchmark and 'Valkyrie,' a scalable infrastructure for running agentic evaluations. By focusing on economically valuable and complex professional tasks rather than academic or contrived problems, Vals AI aims to build the trusted measurement and trust layer for the growing AI economy.

Keywords: AI model evaluation, large language model benchmarks, LLM performance testing, independent AI evaluation, industry-specific AI benchmarks, legal AI, finance AI, healthcare AI, agentic AI testing, AI model scoring, enterprise AI evaluation, Rayan Krishnan, Langston Nashold, Andreessen Horowitz, AI performance metrics, AI trust layer, model evaluation infrastructure, AI auditing, AI certification, SWE-bench, CorpFin benchmark, LegalBench, private AI datasets, AI model accuracy, AI latency testing