TypeSafe AI Launches Jev, a Non-Language AI Model Built for Cheaper, Faster Software Automation
Jev replaces text generation with probability scores, targeting cheaper, more reliable software automation without hallucinations. Vercel saw safety classification run 5–18× faster with greater accuracy, while Bryo AI found it 10–20× cheaper than Gemini.
TypeSafe AI released Jev in September 2026, a transformer-based AI model that forgoes text output entirely in favor of producing probabilities — what the company calls “calibrated decisions.” The launch drew enough developer interest to briefly overwhelm the company’s API.
Jev was built by TypeSafe AI, a startup founded two years ago by Diogo Almeida, a former OpenAI researcher who helped create ChatGPT and co-developed reinforcement learning from human feedback (RLHF). Almeida left OpenAI after concluding that large language models, optimized for human language, were poorly suited for software automation. “Computers speak a different language,” he told TechCrunch.
Unlike standard LLMs, Jev does not generate text. Users define outputs in advance and receive probability scores in return, which the company says eliminates hallucinations. Output tokens are free; input tokens are priced per billion rather than per million, making it significantly cheaper than competing models.
Early adopters report meaningful performance gains. Vercel software engineer Pranit Sharma said replacing OpenAI’s model with Jev for a safety classifier delivered results 5 to 18 times faster with greater accuracy. Bryo AI CTO Nikhil Mudholkar tested Jev against Gemini for email classification and found Gemini slightly more accurate but 10 to 20 times more expensive. He highlighted Jev’s confidence scores as particularly useful for automating workflows.
Beyond classification, developers see potential uses in LLM agent monitoring — using Jev to detect jailbreaks or misbehavior at low cost — and in model routing, where its speed and price make real-time workload sorting practical.
Almeida named the model after 19th-century economist William Stanley Jevons, whose paradox holds that falling commodity costs drive increased use. His goal is to make intelligence cheap enough to embed widely across software. TypeSafe AI says it plans to release additional versions of Jev in new modalities.