Patronus AI has launched a self-serve API platform to detect and prevent AI failures in real time. The platform operates on a pay-as-you-go model, starting at USD 10 per 1,000 API calls for smaller evaluators and USD 20 per 1,000 API calls for larger ones.
The platform's core technology includes Lynx, a hallucination detection model, CopyrightCatcher for detecting protected content reproduction, and FinanceBench for evaluating AI performance on financial questions. The system operates in two modes: a quick-response version for real-time monitoring and a thorough version for deeper analysis. Users can create custom rules in plain English for specific evaluation needs.
The company claims its Lynx model outperforms GPT-4 by 8.3% in detecting medical inaccuracies. The system can identify specific text segments containing hallucinations, allowing engineers to fix problems quickly.
Analyst QuickTake: Patronus AI launched its open-source hallucination detection model Lynx in July 2024 , with the current announcement monetizing this. Previously, the company raised USD 17 million in May 2024 to scale its research, engineering, and sales divisions.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.