TypeSafe AI raises $870M at $7.5B valuation for its non-text AI model Jev

TypeSafe AI has secured $870 million in funding at a $7.5 billion valuation, with Andreessen Horowitz leading the round alongside Sequoia and existing investor DCVC. The deal comes just weeks after the company launched Jev — a model that challenges the dominant paradigm in AI by not generating text at all.

Jev uses a transformer architecture, the same foundation as the large language models powering ChatGPT and Claude, but its output is not prose, code, or conversation. Instead, the model produces probability distributions — what TypeSafe calls "calibrated decisions." Point the model at a business decision, a classification problem, or a workflow step, and it returns a confidence-weighted answer rather than an explanation of one.

Why Not Text?

The pitch is efficiency. TypeSafe claims Jev runs significantly faster and consumes far fewer tokens than comparable LLMs on the same tasks. For enterprises automating high-frequency decisions — routing customer support tickets, flagging transactions, classifying documents — the economics can be compelling. Text generation adds latency and cost that the task doesn't require.

The approach is also more constrained by design. Because Jev doesn't produce open-ended text, there's no freeform output to hallucinate. The model's uncertainty is expressed numerically, which makes it easier to threshold and audit. TypeSafe has positioned this as a feature rather than a limitation: the model tells you how confident it is, and if the confidence is too low, a human can review.

Early Traction

TypeSafe says a third of Fortune 500 companies are already using Jev, a striking adoption rate for a model that launched on September 15 — less than a month before this funding round. The company did not break down how those companies are using it or at what scale.

The founding team brings significant AI research pedigree. Diogo Almeida, previously an OpenAI researcher, co-founded the company alongside Sasha Sheng, a former Meta research engineer, and Erik Gafni, an engineer and entrepreneur. The company was founded in 2024.

Context: AI Beyond Text

The funding reflects growing investor appetite for AI infrastructure that doesn't fit the chatbot mold. As enterprises move past the experimental phase and start deploying AI at scale in production systems, the limitations of text-generating models — latency, cost, unpredictability — become more pressing. Models optimized for decision output rather than language generation represent a different trade-off: narrower, faster, more auditable. Whether that trade-off finds a durable market will depend on whether the use cases that benefit from Jev's approach are large enough to sustain a $7.5 billion company. The Fortune 500 number suggests early momentum is real, as reported by TechCrunch.