Vals AI wants to become the trusted, independent scorecard for model performance as buyers drown in competing lab-published benchmarks.
Vals AI, backed by Andreessen Horowitz, is building out independent benchmarking meant to give enterprises and developers a neutral read on model capability, cutting through marketing-driven leaderboards published by the labs themselves. The pitch is straightforward: as the number of frontier and open models multiplies, buyers need a scorecard they can actually trust.
The company is positioning itself as infrastructure for procurement decisions rather than another model wrapper or app-layer bet.
Every enterprise buying AI right now faces the same problem: labs grade their own homework, and internal eval teams are expensive to staff. An independent, credible benchmark reduces vendor lock-in risk and gives procurement leaders leverage in negotiations with model providers.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →