## What is Jev? "Jev" most commonly refers today to **Jev, a proprietary AI model developed by TypeSafe AI** — a San Francisco–based company founded in 2024. It was released in limited early access on **15 September 2026**, alongside the announcement of a **US$40 million seed round led by DCVC**. ([Wikipedia – Jev (AI model)](https://en.wikipedia.org/wiki/Jev_(AI_model))) ### Key characteristics - **Not a chatbot / not an LLM in the usual sense.** Unlike a large language model, Jev does **not generate natural-language text**. It returns **typed values together with probability estimates and confidence scores**, intended to be consumed directly by other software rather than read by a person. ([Wikipedia](https://en.wikipedia.org/wiki/Jev_(AI_model))) - **"System One model."** TypeSafe describes it as the first of a class of models it calls "System One models" — built to make fast, structured decisions that software can use directly. The company frames it as "a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out." ([TypeSafe AI blog](https://typesafe.ai/blog/introducing-system-one-models-and-jev)) - **Three question types (primitives):** *Choice* (select one option from a defined set, with per-option probabilities), *Score* (rate the state against ordered levels), and *Noul* (evaluate a yes/no statement, returning a probability between 0 and 1). ([Wikipedia](https://en.wikipedia.org/wiki/Jev_(AI_model))) - **No hallucination / no type errors (by design):** Because possible outputs are defined in advance, the model cannot return a value outside the supplied schema. ([TypeSafe AI blog](https://typesafe.ai/blog/introducing-system-one-models-and-jev)) - **Speed & cost claims:** TypeSafe reports end-to-end response times of **70–500 ms**, claiming Jev is **40–200× faster and 40–400× cheaper** than frontier LLMs on comparable tasks. Pricing is listed at **$0.042 per million input tokens**, with output tokens "free (too cheap to meter)." TypeSafe itself notes the workflows were built by its own team, acknowledges possible bias, and says the gains likely sit at the high end of real-world results. ([TypeSafe AI blog](https://typesafe.ai/blog/introducing-system-one-models-and-jev), [Wikipedia](https://en.wikipedia.org/wiki/Jev_(AI_model))) - **Architecture:** TypeSafe says Jev is transformer-based and trained exclusively on synthetic data using a method it calls **Reinforcement Learning for Calibrated Decisions (RLCD)**. However, the company has **not published the exact architecture, weights, or a technical paper**, and outside observers have suggested it may be built on an open-weight LLM. ([Wikipedia](https://en.wikipedia.org/wiki/Jev_(AI_model))) - **Origin:** TypeSafe AI was founded in 2024 by **Diogo Almeida** (CEO, who spent ~4 years at OpenAI working on RLHF, InstructGPT, ChatGPT and GPT-4), **Erik Gafni** and **Sasha Sheng**. Jev was developed in stealth for roughly two years. ([Wikipedia](https://en.wikipedia.org/wiki/Jev_(AI_model))) - **Name:** It is named after 19th-century English economist **William Stanley Jevons**, whose *Jevons paradox* describes how more efficient use of a resource can increase consumption — reflecting the expectation that cheaper machine intelligence will lead to far wider deployment. ([Wikipedia](https://en.wikipedia.org/wiki/Jev_(AI_model))) ### Other meanings of "Jev" - **jev.** — the stage name of **Jephté Kewto**, a Congolese-Canadian rapper who went viral on TikTok for his 2022 single "where's the confetti." ([DuckDuckGo search result summary](https://html.duckduckgo.com/html/?q=Jev%20meaning)) - **Jevons paradox** — the economic phenomenon (efficiency gains increasing rather than reducing total resource consumption) that the AI model is named after. ([Wikipedia – Jevons paradox](https://en.wikipedia.org/wiki/Jevons_paradox)) **Note:** Given the current date (September 2026) and the volume of recent coverage, the AI model is almost certainly the intended referent if you encountered "Jev" in a tech context.