Reid Hoffman and Mark Pincus Back Computer-Use Startup Prentis in Talks for $100M at $1B Valuation

Three months after quietly launching, Prentis is already negotiating a $100 million funding round that would value the computer-use AI lab at $1 billion. The company, co-founded by LinkedIn's Reid Hoffman and Zynga's Mark Pincus alongside CEO Ritankar Das, emerged from stealth in April 2026 and is moving fast to capitalize on investor appetite for AI agents that can operate software the way humans do.

Das, who leads Prentis, is also the founder of Titan, a holding company that incubates and runs AI ventures. He dropped out of an AI PhD program at Cambridge, where he held a Gates Cambridge Scholarship, and previously graduated from UC Berkeley at 18 as its youngest University Medalist in more than a century, with a double major in bioengineering and chemical biology. He also holds a master's in biomedical engineering from Oxford.

Prentis is betting that "computer use" models, systems trained to navigate desktop and web interfaces autonomously, represent the next major frontier in enterprise AI. That thesis has already drawn a heavyweight competitor into the space: Anthropic acquired Seattle-based computer-use startup Vercept earlier in 2026, absorbing its founders and discontinuing its product.

Early revenue signals

Unlike many labs still chasing product-market fit, Prentis says it has already signed up to $50 million in contracts with customers spanning a healthcare management service organization, a manufacturer, and clothing and goods producers. Investor materials reviewed by TechCrunch project an annualized run rate of roughly $75 million by the third quarter of 2026.

The company's core model, Hive-32B, is central to its pitch. According to Prentis's own deck, it outperforms OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.6 on the WindowsAgentArena and ScreenSpot-v2 benchmarks, both standard tests for how well AI systems can operate graphical interfaces.

Prentis claims Hive-32B delivers roughly 10 times lower cost per task than frontier APIs, a figure central to its pitch for enterprise adoption.

That cost advantage, if it holds up under independent scrutiny, could be the more consequential claim than the benchmark wins themselves. Enterprises evaluating computer-use agents care as much about per-task economics at scale as raw accuracy, and a tenfold cost gap against GPT-5.4 or Claude Opus 4.6 would reshape procurement conversations regardless of who tops a leaderboard.

Minor inconsistencies appear across reporting on the deal, including whether Mira Murati's lab is referred to as "Thinking Machines Lab" or simply "Thinking Machines," and slight variation in the spelling of Pincus's first name. Neither affects the substance of Prentis's fundraising story, but they underscore how fast-moving and lightly verified some of the early coverage around this round has been.

Prentis has not disclosed which investors are leading the talks, and the $100 million round has not yet closed. If finalized at the reported $1 billion valuation, it would mark one of the more aggressive markups for a computer-use specialist just months removed from stealth.