TypeSafe AI, the startup behind the Jev model, has raised $870 million in a Series A round that values the company at $7.5 billion. Andreessen Horowitz led the financing, with Sequoia Capital, existing investor DCVC and a group of unnamed angel investors taking part, and a16z general partner Martin Casado is joining TypeSafe's board. The company disclosed the round in a short blog post on October 9, the same day TechCrunch and SiliconANGLE reported it and a16z published its own investment note.
The size of the round is striking mainly because of how young the product is. Jev was released on September 15, so the funding arrived roughly three and a half weeks after launch. TypeSafe was founded in 2024 by Diogo Almeida, a former OpenAI researcher who is now chief executive, former Meta research engineer Sasha Sheng, and engineer and entrepreneur Erik Gafni.
Jev is unusual among headline-grabbing models in that it does not produce text. It is built on a transformer architecture but is not a large language model; instead of writing sentences or code, it returns probabilities that the company describes as calibrated decisions, meant to be called directly from software. TypeSafe pitches this as a better fit for automation than general-purpose chatbots, arguing that computers do not need human language to make routine choices. The company says Jev runs much faster and consumes far fewer tokens than language models doing comparable work, claims that have not been independently benchmarked.
Adoption figures come from the company and its new lead investor, and they do not match exactly. TypeSafe's post says a third of the Fortune 500 are using Jev and that it has already saved customers millions of dollars in production. A16z's announcement puts enterprise integration at 25 percent of the Fortune 500 and says Jev reached one trillion generated tokens in its first three days, calling it the fastest-growing model the firm has seen. A16z frames Jev as the first of a class of "System One" models designed to be embedded throughout software at costs closer to traditional computing.
TypeSafe's post is light on specifics about how the money will be spent. It promises more machine-native models, broader infrastructure for building what it calls smart software, and the enterprise features customers have requested. SiliconANGLE reported that Jev is part of a planned System One model lineup that the proceeds will help expand.
For the wider market, the round is a large bet that a meaningful share of the work businesses now send to language models is really decision-making that can be handled more cheaply by a model that skips text entirely. If that holds, TypeSafe has an early lead in a category with few established rivals. If general-purpose models keep getting cheaper at the same tasks, the advantage could narrow quickly.
Several questions remain open. Neither TypeSafe nor its investors have published independent performance evaluations, customer names or revenue, and the two adoption figures differ. The company has not said whether the valuation is pre- or post-money. With the round coming so soon after launch, the valuation rests far more on early usage momentum and investor conviction than on a long commercial record.