A new AI model from a ChatGPT inventor is challenging the assumption that all useful AI must work with language. Jev, released this week by TypeSafe AI, is not a large language model (LLM) at all. Instead of generating text like ChatGPT or Claude, it produces calibrated probabilities and decisions, making it radically cheaper, faster, and impossible to hallucinate.

Why Would an AI Skip Language Altogether?

The creator, Almeida, was an OpenAI researcher who helped build ChatGPT and invented reinforcement learning from human feedback (RLHF), the training technique that made today's conversational AI possible. Yet despite those achievements, he felt something was missing. "We have lightning in a bottle, and yet it is not useful," Almeida told TechCrunch. "I've been battling that problem since then. It took me a while to come to the conclusion: the problem is we are optimizing for human language. We have been super good at human language for four years, but it's not useful for automation because computers speak a different language."

Two years ago, Almeida left OpenAI to start TypeSafe AI with a different vision. Instead of chasing ever-larger language models, the company built Jev as what it calls a "System One model," focused on intuition and fast decisions rather than reasoning. The model is trained exclusively on synthetic data using a technique Almeida calls "reinforcement learning from calibrated decisions."

How Much Faster and Cheaper Is Jev Than ChatGPT?

The practical differences are striking. Because Jev outputs probabilities rather than text tokens, its output tokens are free, and input tokens are metered by the billion instead of the million. This cost structure makes it possible to use AI for tasks that would be prohibitively expensive with ChatGPT or other LLMs.

Real-world tests show the speed advantage. Pranit Sharma, a software engineer at Vercel, a company building agentic infrastructure, tested Jev against OpenAI's ChatGPT Luna 5.6 for a safety classifier. When Vercel switched from ChatGPT Luna to Jev, it got results 5 to 18 times more quickly and with greater accuracy. In another test, Bryo AI's CTO Nikhil Mudholkar compared Jev against Google's Gemini for classifying business emails. Gemini was slightly more accurate, but Jev was 10 to 20 times less expensive. More importantly, Mudholkar noted that Jev was "the only one that hands back a real probability which makes it ideal for automating workflows."

Mudholkar

The demand has been so high that TypeSafe AI briefly lost the ability to serve users from its API. Developers are clearly hungry for a different approach to AI integration.

What Can You Actually Do With a Non-Language AI Model?

Jev appears most useful for software automation, where developers need fast, reliable decisions rather than creative text generation. The model opens up several practical applications that were previously too expensive or slow to implement:

• Replacing LLMs in specific tasks: For classification, routing, and decision-making, Jev can do the job faster and cheaper than ChatGPT or other language models.
• Monitoring other AI agents: Almeida suggests using Jev to track LLM agent traces and prevent jailbreaks, acting as a smart check on misbehavior without the cost of using another LLM to monitor the first one.
• Model routing in real time: Predicting whether a workload requires a specific model would normally be expensive if done with an LLM, but Jev's low cost and speed make real-time sorting possible.
• Confidence-based decision making: Because Jev returns probability scores, developers can set thresholds for action, disregarding low-confidence results and acting on high-confidence ones.

Armin Ronacher, the CTO of Earendil, which builds the open-source model harness Pi, explained the practical value: "At the end of the day, it delegates the hallucination problem a little bit to the user. The user has to say, okay, if this only comes back with 50% probability, maybe this is a coin toss, and I disregard it. But if it's 95%, sure, then I can do something with it."

What's the Bigger Vision Behind Jev?

The model is named after William Stanley Jevons, a 19th-century economist whose paradox describes how falling costs lead to increased use of a commodity. Almeida hopes the same principle applies to AI: as the cost of intelligence drops, it should become embedded everywhere in software, not just in high-stakes applications. "We think that there's just going to be smart software all over the place in a way that's emergent and distributed, much more like the early internet than you know like the mega apps that people are trying to build right now," he said.

Almeida is tight-lipped about Jev's architecture, which outside observers suspect is built on top of an open-weight LLM. He emphasized that the company's biggest bet was on synthetic data. "We made an early bet that we will be making all of our data, and that has been one of the best bets I've ever made in my life, better than our launch, in my opinion, better than RLHF," he told TechCrunch. "Half of our company is a lab that basically owns this entire subfield of statistically well-understood synthetic data, and that is now my life joy."

For now, Jev stands alone as this kind of model, but competition is likely coming. Ronacher noted that "we should have seen this earlier in many ways, but presumably because the LLMs are so cheap and subsidized, you often don't have to be creative yet." As the economics of AI shift, other companies may follow TypeSafe AI's lead and build models optimized for specific tasks rather than general conversation.

Ronacher

The broader implication is that the era of one-size-fits-all large language models may be giving way to a more specialized landscape, where different AI tools are optimized for different jobs. For developers building automation and intelligent software, Jev represents a fundamentally different approach to integrating AI into their systems.