TypeSafe AI announced on October 9, 2026 that it raised an $870 million Series A at a $7.5 billion valuation, led by Andreessen Horowitz with participation from Sequoia Capital, DCVC and angel investors. Andreessen Horowitz general partner Martin Casado joins the board. The round follows the September 15 early access launch of Jev after two years in stealth, and it matters because the company positions Jev as a separate model class for automated decisions inside software.
Funding round and Jev early access launch
Andreessen Horowitz confirmed the lead investment in an October 9 post bylined by Jennifer Li, Sarah Wang, Martin Casado, Marc Andreessen and Ben Horowitz. TypeSafe disclosed the amount, valuation and board seat in a footnote to a short post addressed to developers, businesses and recruits. The company said it is bringing developers off its waitlist and plans to ship additional machine-native models, add infrastructure for smart software and add requested enterprise features. The announcement also links to its careers page for hiring.
Jev takes typed questions, evaluates them against a supplied state and returns structured results without generating or parsing text. Outputs arrive as typed values with probability distributions that code can branch, sort and route directly. Three primitives are exposed: Choice selects one option from a list, Score grades a state against a rubric, and Noul returns a 0-1 answer on whether a statement is true. Choice and Score also return confidence values, and all three can be combined in one API call with independent concurrent evaluation.
Founder Diogo Almeida described Jev as the first entry in a new class called System One, built for rapid structured decision-making. The stack includes a new model architecture, a parallel sampler and a training method called Reinforcement Learning for Calibrated Decisions. The company reports 70 to 500 millisecond end-to-end response times, described as 40 to 200 times faster than frontier models at comparable intelligence on System One-shaped queries. Because outputs are predefined type-safe values, the model never makes type errors and cannot hallucinate, according to the post.
What faster classification means for business
Published pricing is $0.042 per million input tokens, with output tokens free. Andreessen Horowitz characterized Jev as roughly 1/100 to 1/500 the cost of frontier models while running 100 times faster for classification tasks at comparable accuracy. TypeSafe notes that homepage claims of 193.6-times-faster and 444.6-times-cheaper come from its own workflow evaluations, sit at the higher end of real-world gains and may contain bias because the workflows were built by its model capabilities team. Adoption claims differ: TypeSafe reports a third of the Fortune 500 using Jev, while the investor post cites 25% integration.
The investor post called Jev the fastest-growing model it has seen, with 1 trillion tokens generated within three days of launch and thousands of use cases in the first week, including generative UI, interactive gaming and data analysis. For operating companies, the practical effect is cheaper high-volume screening, routing and scoring passed directly into code as typed values. An October 7 case study said talent marketplace Jack & Jill replaced Gemini 3.1 Flash Lite for 100% of calls in a key matching stage within 10 days, cutting screening cost by 88% from $0.755 to $0.092 per 1,000 candidates.
The same case study reported median screening time falling from 20.3 to 10.3 seconds, with ranking quality at 0.933 AUC for Jev versus 0.924 for Gemini and retention of 94.6% of candidates requested for meetings against a 93.9% baseline, a difference described as not statistically significant. Reported savings were $265,000 annually at current volume and $500,000 projected over twelve months, with use expanded to more than 15 workflows. Before committing, buyers should verify savings on their own traffic, test calibration of confidence values and confirm latency under production load.
