The maker of non-text AI model Jev valued at $7.5B just weeks after launch

TypeSafe AI, the startup behind the Jev model, has raised $870 million at a $7.5 billion valuation, TechCrunch reports. Andreessen Horowitz led the round, with Sequoia and existing investor DCVC participating. Jev only launched on September 15.

Jev is built on a transformer architecture but is not a large language model. It does not output text. It produces probabilities, which the company calls "calibrated decisions." TypeSafe claims it is significantly faster and uses far fewer tokens than LLMs, and pitches it for automating tasks rather than writing text or code. The startup says a third of Fortune 500 companies already use it.

If that holds up, it points to a quieter but huge part of AI: the yes-or-no, pick-one, route-this calls that businesses make millions of times a day.

We like the idea underneath this, even as we raise an eyebrow at the speed of the money. A lot of enterprise AI today is a language model asked to make a simple call, then a parser trying to pull a label out of its prose. That is slow, expensive and brittle. A model that answers the question directly, with a probability attached, is a cleaner fit for automation. Co-founder Diogo Almeida's line that computers speak a different language than people gets at something real.

Probabilities are also more honest than confident paragraphs. A system that says it is 62 percent sure invites a threshold and a human check. A system that writes a fluent sentence invites blind trust. If Jev's outputs are genuinely calibrated, that is a safety feature as much as a speed feature.

Now the caution. The adoption number is the company's own claim, and "using" can mean anything from a pilot to production. Calibration is easy to promise and hard to prove across messy real data. We have not seen independent evidence in this report either way. And a $7.5 billion valuation weeks after launch prices in a lot of success before the market has really tested it.

Competition is also arriving fast. OpenAI has just put its own typed-answer endpoint into public beta, which tells us the big labs see the same opportunity. TypeSafe's edge will need to be price, speed and trust, not novelty.

The downside if this scales is the one every automation wave brings. Faster, cheaper decisions mean more decisions made without a person in the loop, in claims, moderation, routing and screening. That is useful when the stakes are low and the thresholds are sensible. It is harmful when a probability quietly becomes a denial. I'm watching for whether customers publish how they set those thresholds.

TypeSafe was co-founded in 2024 by Diogo Almeida, previously a researcher at OpenAI, former Meta research engineer Sasha Sheng, and engineer and entrepreneur Erik Gafni. Jev went viral almost immediately after its September 15 release.