Reid Hoffman and Marc Pincus Back AI Agent Startup Prentis

The company, which launched in April, is developing models designed to navigate routine administrative tasks such as processing insurance claims and managing customs refunds without human intervention. Early traction appears significant, with the startup reportedly securing contracts valued at up to $50 million. Pitch materials suggest these agreements could translate to a $75 million annualized run rate by the third quarter, though these figures remain dependent on finalized performance metrics.

Technically, Prentis distinguishes itself by prioritizing smaller, more cost-efficient models. It claims its proprietary Hive-32B architecture outperforms larger industry rivals like GPT-5.4 and Claude Opus 4.6 on specific benchmarks, including WindowsAgentArena and ScreenSpot-v2. By reducing the cost per task by roughly tenfold compared to frontier APIs, the firm intends to capture a market segment currently crowded by major players like Anthropic and OpenAI. With a team of 25 employees drawn from firms like Google DeepMind and Meta, the startup is positioning computer-use automation as the next major shift in artificial intelligence, potentially eclipsing current coding-focused applications.