Openbenchmarks

Independent benchmarks for AI agents choosing tools

Openbenchmarks builds domain-specific, reproducible evaluations that AI agents can use to decide which tools to pick. Agents now research products, compare options and even make build-versus-buy calls for the people they work for, and they favour benchmarks that are open, independent and grounded in data, so any company that wants agents to choose its product needs to show up in credible evaluations. The founders previously led AI research and infrastructure teams at Oracle and AppFolio, and the company grew out of their research into model behaviour and tool selection. It works with AI-first companies such as Parallel, Firecrawl, Telnyx and TinyFish, and writes up its results for human readers and for the agents that index and cite them.