Jev AI Model Cuts Decision Costs for Developers

TypeSafe has launched Jev, a specialised AI model built for fast, structured decisions rather than writing, explanation or long-form text generation. Early reports said Jev delivered up to 193.6 times higher speed and 444.6 times lower cost than larger models in selected tests. TypeSafe later reported that around 13% of Vercel’s paid teams adopted Jev within 24 hours, while Cloudflare, LangChain and Langfuse added native support within days. Jev handles binary decisions, multiple-choice tasks, scoring and calibrated probability estimates. TypeSafe claims it is 40 to 200 times faster and 40 to 400 times cheaper than comparable large language models. Pricing starts at $0.042 per million input tokens, while output tokens are free. Independent tests found five- to 18-fold speed gains in a Vercel security classifier and costs 10 to 20 times below Gemini in commercial email classification, although Jev was slightly less accurate. In an agent simulation, Jev completed 13,200 decisions for $0.35, compared with an estimated $37.64 using frontier models. TypeSafe was founded by former OpenAI researcher Diogo Almeida and raised a $40 million seed round led by DCVC. Its RLCD method, or Reinforcement Learning for Calibrated Decisions, is intended to improve the reliability of confidence scores. Jev is designed to manage high-volume AI agent tasks such as tool selection, context filtering, webpage actions and completion checks, while larger models handle complex reasoning. Jev remains limited in mathematics, counting, date comparisons and multi-step logic. Its architecture is undisclosed, and community developers are working on projects such as OpenJev. For crypto traders, Jev highlights the expanding market for specialised AI inference and cheaper automation, but limited production data and uncertain commercial sustainability suggest little immediate impact on cryptocurrency prices.