San Francisco startup TypeSafe AI has released Jev, a new artificial intelligence model designed to return typed values and probabilities based on predefined output schemas.
San Francisco startup TypeSafe AI has released a new artificial intelligence model called Jev. Rather than generating traditional human text, Jev returns typed values and probabilities based on a predefined output schema.
What Happened
Released last week by TypeSafe AI—a company founded in 2024 by Diogo Almeida, Erik Gafni, and Sasha Sheng—the Jev model quickly drew intense interest from the developer community. Demand for the model was so high that the startup's API temporarily stopped serving requests.
Unlike standard large language models that produce text one token at a time, Jev operates on a different mechanism. Developers establish an output schema beforehand, outlining the exact structure of every potential answer, before feeding in unstructured state data such as support tickets.
Key Details
- Company: TypeSafe AI (founded in 2024 by Diogo Almeida, Erik Gafni, and Sasha Sheng)
- Funding: Raised a $40 million seed round led by DCVC
- Response Times: End-to-end response times range from 70 to 500 milliseconds by generating outputs in parallel within a single query
- Pricing: Input costs are $0.042 per million tokens, while output tokens are free
- Output Type: Returns typed values with attached probabilities, which TypeSafe calls 'calibrated decisions' and a 'frontier-intelligence function call'
Why It Matters
The model's unique approach to handling inputs and producing structured outputs offers developers a distinct alternative to traditional generative text models. By generating parallel outputs within a single query, Jev achieves rapid end-to-end response times.
What We Know So Far
Confirmed facts include the company's founding details, funding amount led by DCVC, technical pricing, and performance metrics reported by TypeSafe AI. An unconfirmed report from Forbes, summarized by SiliconANGLE, places the company's valuation at about $200 million.
