here is a new AI model getting a lot of attention from developers this week, and it does something almost no other AI model does. It never writes a single sentence. It is called Jev, and it comes from a company called TypeSafe AI. Here is exactly what it is and why it matters, explained simply.
Quick Answer
- Jev is a new AI model from a company called TypeSafe AI, launched September 15, 2026
- Unlike ChatGPT or Gemini, Jev does not write text at all
- It only returns decisions, like a choice, a score, or a yes/no answer, each with a confidence percentage
- It can be up to 100 times faster and cheaper than regular AI models for these kinds of tasks
- Because its answers are limited to a fixed set of choices, TypeSafe says it cannot hallucinate
- It was built by Diogo Almeida, a former OpenAI researcher who worked on the technology behind ChatGPT
- It is currently in early access, behind a waitlist
What Is Jev AI
Jev is TypeSafe AI’s first model, and it belongs to a completely new category the company is calling a System One model. The name is a reference to a famous idea from psychology, the difference between fast, gut-feeling thinking (System 1) and slow, careful reasoning (System 2). Most AI chatbots today are built to reason slowly and explain themselves. Jev is built to do the opposite: make a fast, confident decision and move on.
Instead of asking Jev an open-ended question, you give it some information, called the state, along with a specific question that has a limited set of possible answers. Jev responds almost instantly with its answer and a confidence score, like “block this transaction: 92% confidence,” instead of writing out a paragraph explaining itself.
Why Is Jev Different From ChatGPT or Gemini
Regular AI chatbots like ChatGPT, Claude, and Gemini generate text one word at a time, which takes time and costs money for every word produced. Jev skips that process entirely. It was not built to chat with people. It was built to sit quietly inside software and make small, structured decisions, the kind of thing apps already need constantly: should this message be flagged as spam, is this the right response, should this action be allowed or blocked.
A lot of companies were already using expensive chat-based AI models just to answer simple yes-or-no or multiple-choice questions inside their software. Jev is built specifically to replace that use case at a fraction of the cost and time.
How Fast and Cheap Is Jev
This is the part that has gotten developers excited. TypeSafe says Jev can respond in under 100 milliseconds, and independent tests have shown it running dramatically faster than typical AI models on structured tasks, with some comparisons showing it around 25 to 100 times faster, and far cheaper per task, sometimes hundreds of times cheaper depending on the comparison.
On pricing, Jev charges a small fee for the information you send it, and does not charge anything at all for its answers. That is unusual. Most AI models charge for both what you send and what you get back.
Does Jev Really Never Hallucinate
TypeSafe’s claim is that Jev cannot hallucinate, and the reason is architectural, not just good training. Since developers define the exact list of possible answers ahead of time, Jev physically cannot return something outside that list. It cannot make up a fact or invent an answer that was never one of the allowed options. This is very different from a chatbot, which can generate any text at all, including wrong or made-up information.
Who Made Jev
Jev comes from TypeSafe AI, a San Francisco based startup that came out of stealth mode on the same day it launched Jev. The company’s founder and CEO is Diogo Almeida, who previously worked at OpenAI and Google Brain. He is credited by his own company as a co-inventor of the technology behind ChatGPT, though it is worth noting he was one of several primary researchers on that original work, not the sole creator. He also contributed to GPT-4 during his time at OpenAI.
TypeSafe was co-founded alongside Erik Gafni and Sasha Sheng, and the company raised a $40 million seed round led by investment firm DCVC.
What Can Jev Actually Be Used For
Jev is not meant to replace chat-based AI models like Meta’s Muse or tools like ChatGPT and GPT-6 Astra for open-ended conversations or writing. It is meant to work alongside them. Some real early examples:
- Choosing which AI model should handle a task. A cheap, fast decision layer like Jev can look at a request first and route it to a bigger, smarter model only when it is actually needed.
- Checking AI agent actions before they happen. Jev can act as a safety check, deciding whether an action an AI agent is about to take looks risky, before it is allowed to proceed.
- Content and email classification. Sorting messages, flagging spam, or categorizing support tickets without needing a full chatbot model for the job.
- Real-time systems. Early demos from TypeSafe included things like a game-playing bot and a simulated self-driving scenario, situations where waiting several seconds for a chatbot-style response simply does not work.
Is Jev Available to Try Right Now
Jev is currently in early access. You need to join a waitlist through TypeSafe AI to get access, so it is not yet something the general public can use freely. It has mostly been picked up by developers and companies building AI-powered software rather than everyday consumers.
Frequently Asked Questions
What is Jev AI?
Jev is a new type of AI model from TypeSafe AI that returns structured decisions, like a choice or a score, instead of writing text like a chatbot does.
Who created Jev?
Jev was created by TypeSafe AI, a startup founded by Diogo Almeida, a former OpenAI and Google Brain researcher who worked on the technology behind ChatGPT.
How is Jev different from ChatGPT?
ChatGPT generates open-ended text, one word at a time. Jev only picks from a fixed set of possible answers you define in advance, and returns its answer almost instantly with a confidence score.
Is Jev faster than regular AI models?
Yes. TypeSafe and independent testers have measured Jev as significantly faster, often tens to hundreds of times faster, than typical AI models on structured decision tasks.
Can Jev make things up like a chatbot sometimes does?
TypeSafe says no, because Jev can only choose from a predefined list of answers, it cannot generate a made-up response the way open-ended chat models sometimes can.
How much does Jev cost to use?
Jev charges a small fee for the information sent to it and does not charge for its responses, making it much cheaper than typical AI models for high-volume, simple decisions.
Can I use Jev right now?
Jev is currently in early access through a waitlist on TypeSafe AI’s website. It is mainly being used by developers and companies right now, not general consumers.
