Patronus

Patronus AI is a frontier AI research lab and infrastructure company that develops Digital World Models to simulate and predict agent actions within digital workflows. The company has trained the first Digital World Model, which generates interactive digital worlds dynamically for AI agents, achieving 30-40% model lift on long-horizon tasks. The platform has accumulated over 1 million world data artifacts across diverse domains with 85% UI/UX feature parity with real-world products, supported by over 5,000 expert contributors across software, academia, finance, and other fields. Patronus AI's research includes Lynx, a state-of-the-art hallucination detection model that outperforms GPT-4; FinanceBench, an industry-first benchmark for LLM performance on financial questions; BLUR for evaluating agent effectiveness; and GLIDER, an evaluation model for explainable reasoning chains. Simulation capabilities span deep research, multi-turn dialogue, long-horizon task planning, and agentic memory. Founded in 2023 and headquartered in New York City, Patronus AI has raised $70.0M in Series B funding from investors including Lightspeed Venture Partners, Factorial, Notable Capital, Datadog, Greenfield Partners, and Samsung.

AI Infrastructure > Machine Intelligence Systems > Artificial General Intelligence

Company Wise > Facebook,

College Wise > Stanford University, University of Chicago