Vals, backed by Andreessen Horowitz, is looking to become the gold standard for AI benchmarking
techcrunch ·
Vals, a startup founded in 2024, wants to overhaul AI benchmarking, which companies use to validate model capabilities and promote their advantages, even though many established tests are outdated and can be gamed. After a seed round led by 8VC and Bloomberg Beta, it raised a $40 million Series A led by Andreessen Horowitz last month following rapid growth. Co-founder Rayan Krishnan says Vals keeps specific test materials private and focuses on whether models can complete complex, industry-specific work in areas such as law, finance, and coding rather than only testing general knowledge. The company also examines both beneficial and harmful potential effects, exploring areas including recursive self-improvement, mental health, cybersecurity, biosecurity, and the law of armed conflict. Companies pay for these evaluations because they can help diagnose and improve models, and the results are increasingly influencing decisions about which AI models to acquire. Vals says its revenue is eight times last year’s level; its team has grown from eight to 25, it plans to move to a larger office and hire 10–15 more people, and it has launched an evaluation program for federal agencies. Krishnan believes such benchmarks will become central to AI companies’ public trust, usage, filings, and investment discussions as their models spread through the economy.
techcrunch