TypeSafe AI announced on October 9, 2026 that it has raised an $870 million Series A at a $7.5 billion valuation, led by Andreessen Horowitz with participation from Sequoia Capital, existing investor DCVC, and angel investors. Andreessen Horowitz general partner Martin Casado is joining the company’s board.
Round Terms and Investors
Andreessen Horowitz confirmed it is leading the investment in a post dated October 9, 2026 and bylined by Jennifer Li, Sarah Wang, Martin Casado, Marc Andreessen, and Ben Horowitz. Li and Casado are general partners focused on the firm’s infrastructure investing, Wang is a general partner on its growth team, and Andreessen and Horowitz are the firm’s cofounders and general partners.
TypeSafe’s own announcement is a short post addressed to developers, businesses, and potential recruits; the company disclosed the dollar amount, the valuation, and Casado’s board seat in a footnote to that post.
The System One Model Behind the Round
TypeSafe released Jev in early access on September 15, 2026, after two years in stealth, and said it is bringing developers off its waitlist. In the launch post, founder Diogo Almeida described Jev as the first entry in a new model class the company calls System One, designed for rapid, structured decision-making inside software.
According to TypeSafe’s documentation, Jev takes typed questions and evaluates them against a supplied state, returning structured results without generating or parsing text. The results arrive as typed values with probability distributions, so code can branch, sort, and route on them directly. The company exposes three primitives: Choice picks one option out of a provided list; Score grades the state against a rubric; and Noul returns a 0–1 answer to whether a given statement is true. Choice and Score also return confidence values, and all three question types can be combined in one API call, where each is evaluated independently and concurrently against the same state.
Almeida wrote that TypeSafe built a new stack focused entirely on automation: a new model architecture, a parallel sampler, and a training method the company calls Reinforcement Learning for Calibrated Decisions (RLCD). He said Jev achieves similar intelligence to existing large language models on System One tasks while running two orders of magnitude faster, with end-to-end response times of 70 to 500 milliseconds, a range the post states can be 40 to 200 times faster than frontier models at comparable intelligence for System One–shaped queries. Because outputs are type-safe structured values defined in advance, Almeida wrote, the model never makes type errors and cannot hallucinate.
Published pricing is $0.042 per million input tokens, with output tokens free. The company states that the 193.6-times-faster and 444.6-times-cheaper figures on its homepage come from its workflow evaluations, and it cautions that those numbers are on the higher end of real-world gains. It also notes that the evaluation workflows were built by members of its model capabilities team and that some bias could exist.
Almeida wrote that at OpenAI he helped develop instruction-following methods for language models, work he said became the research behind ChatGPT. The System One name references Daniel Kahneman’s book Thinking, Fast and Slow and its contrast between quick, intuitive System 1 thinking and slower, deliberate System 2 reasoning. Jev is named after William Stanley Jevons; the company said it expects machine intelligence to follow the path of coal, for which steam-engine efficiency gains led to increased demand.
Reported Adoption and Early Deployment
TypeSafe said in its announcement that a third of the Fortune 500 are using Jev and that it has already saved customers millions of dollars in production. The Andreessen Horowitz post states a different figure, saying 25% of Fortune 500 enterprises have integrated Jev. The firm described Jev as the fastest-growing model it has seen, stating that it reached 1 trillion tokens generated within three days of launch and that thousands of use cases surfaced during the first week, among them generative UI, interactive gaming, and data analysis. It characterized Jev as roughly 1/100 to 1/500 the cost of frontier models while being 100 times faster for classification tasks at comparable accuracy, passing model decisions straight into code as typed values rather than text that software must then parse.
In an October 7, 2026 case study, TypeSafe reported that Jack & Jill, a talent marketplace where AI agents match candidates with hiring teams, replaced Gemini 3.1 Flash Lite for 100% of calls in a key stage of its candidate-matching pipeline within 10 days of its first test. According to the case study, the change cut screening costs by 88%, from $0.755 to $0.092 per 1,000 candidates, and halved median screening time from 20.3 seconds to 10.3 seconds, while retaining 94.6% of candidates that hiring managers later asked to meet, compared with a 93.9% baseline; the study notes that this quality difference was not statistically significant. Ranking quality measured 0.933 for Jev versus 0.924 for Gemini on AUC. TypeSafe reported $265,000 in annual savings for Jack & Jill at current volume, with $500,000 projected over the next twelve months given expected growth, and said Jev usage has expanded to more than 15 workflows, including matched roles shown to candidates at sign-up and a profile check that fell from about 3.6 seconds to about 0.2 seconds.
Stated Plans
TypeSafe said it intends to take Jev’s capabilities further, ship additional machine-native models, provide infrastructure for building smart software, and add the enterprise features customers have requested. The announcement also encourages potential recruits to join, linking to the company’s careers page.
Credit: Source link
