TypeSafe Raises $870 Million at $7.5 Billion 24 Days After Jev's Launch, and Its Fortune 500 Figure Has Two Versions

TypeSafe AI, the maker of the Jev model, raised $870 million at a $7.5 billion valuation on Oct. 9, 24 days after Jev's early-access launch on Sept. 15, TechCrunch reported. Andreessen Horowitz led, with Sequoia Capital and existing investor DCVC.

The headline valuation is the least informative number here, because no revenue figure has been published to set it against. What can be checked is the pricing and the adoption claims, and the adoption claims disagree.

Two Fortune 500 numbers

SiliconANGLE and TechCrunch report TypeSafe's claim that about a third of the Fortune 500 use Jev. Unite.AI, summarising a16z's announcement, says a16z puts it at 25%. The Terminal could not retrieve the a16z post itself, so that second figure rests on Unite.AI, which labels its byline an AI-generated research agent. No source names a customer.

Claim	TypeSafe	a16z, via Unite.AI
Fortune 500 using Jev	about one third	25%
Speed vs frontier LLMs	up to 200x	100x for classification
Cost vs frontier LLMs	up to 100x cheaper	1/100 to 1/500

What Jev is, and what it costs

Jev is transformer-based but not a language model. It takes a typed question, evaluates it against supplied state and returns a typed value with a confidence number: a Choice from a list, a Score against a rubric, or a 0 to 1 truth value. It is the first of a planned System One series, per SiliconANGLE, and was trained with a method TypeSafe calls Reinforcement Learning for Calibrated Decisions. Latency is under 700 milliseconds, which TypeSafe says is up to 200 times faster than some frontier models. Those are vendor-supplied numbers.

Input tokens cost $0.042 per million, and output tokens are free, according to Unite.AI. That is the arithmetic worth doing. A customer base paying $100 million a year at that rate sends about 2.4 trillion input tokens. a16z says Jev generated 1 trillion tokens within three days of launch, but those are generated tokens, which are free, so they say nothing about billed volume. At $7.5 billion, $100 million of revenue would be a multiple of 75, and $100 million is a hypothetical, not a disclosed figure.

One customer case, vendor-published

The only named customer data is a case study TypeSafe published on Oct. 7, as summarised by Unite.AI. Jack & Jill, a talent marketplace, replaced Gemini 3.1 Flash Lite for 100% of calls in one candidate-matching stage within 10 days. Cost per 1,000 candidates fell 88%, from $0.755 to $0.092. Median screening time went from 20.3 to 10.3 seconds. Quality was level: 94.6% of candidates later requested by hiring managers were retained against 93.9% on the baseline, a gap the study itself rates as statistically indistinguishable, and AUC was 0.933 against 0.924.

1. Gemini 3.1 Flash Lite0.76 $
2. Jev0.09 $

Source: TypeSafe case study dated Oct. 7, 2026, as summarised by Unite.AI (vendor-supplied)

The saving is $265,000 a year at current volume, per the case study. One $265,000 saving does not explain a $7.5 billion price, and the company says it plans more System One models and enterprise features with the money.

For scale, Oratomic's $475 million Series B is the other nine-figure round covered here this week. The category is not new either. Liquid AI opened d1-3B, a decision model, with its own scoreboard. The next fact that would settle this is a revenue figure or a named Fortune 500 customer. Waitlist access for developers is opening, per Unite.AI, so independent latency and cost tests could follow.