TypeSafe Hits 7.5B Valuation as Jev Decision Models Boom

TypeSafe AI, the startup behind the non-text Jev model, has raised $870 million at a $7.5 billion valuation just weeks after launch, a sign that the "decision model" category is becoming one of the fastest-moving corners of AI.

What happened

The round was led by Andreessen Horowitz, with Sequoia and existing investor DCVC participating, according to TechCrunch. TypeSafe was founded in 2024 by Diogo Almeida, a former OpenAI researcher, alongside ex-Meta research engineer Sasha Sheng and engineer-entrepreneur Erik Gafni.

Jev, released on September 15, is a transformer but not a large language model. Instead of generating text, it takes unstructured program state plus a set of typed questions and returns calibrated decisions - a choice, a score, or a yes/no answer with a confidence value. TypeSafe says the model runs significantly faster and uses far fewer tokens than an LLM, which it pitches as uniquely suited to automating tasks rather than writing prose.

Adoption has been unusually quick. TypeSafe claims a third of Fortune 500 companies are already using the model, an unusually swift enterprise ramp for a product that is barely a month old.

Why it matters

The raise caps a month in which decision models went from a novelty to a crowded field. OpenAI shipped a Jev-style offering in late September, and Amazon open-sourced its own take, Strands Decider 2B, on October 1.

Amazon distinguished engineer Marc Brooker told TechCrunch the appeal is a decider for a workflow step: finding the next action based on current state. He said the confidence scores and closed domain of answers make such a step more reliable, with lower latency and potentially lower cost than calling a frontier LLM every time.

Builder impact

For teams building agents and automations, the decision-model wave reframes a familiar design choice. Not every step in a workflow needs a general-purpose model. Routing, classification, and pick-the-next-action steps can be handled by a small decision model that returns a typed answer with a confidence threshold: act when confidence is high, escalate when it is low.

That is exactly where agent builders spend the most tokens today. If a cheap, fast, locally runnable decider can absorb those steps, the cost curve for multi-step agent workflows changes. The open question, as Brooker notes, is whether small decision models can stay accurate without losing the general knowledge that makes them useful.

With $870 million in fresh capital, TypeSafe now has to prove the category is more than a gold rush. Its executives told TechCrunch they are keeping their heads down and improving future models.