Ex-OpenAI researcher's non-LLM model stuns developers

TypeSafe AI, founded by ex-OpenAI researcher Diogo Almeida, has released Jev, a transformer-based model that outputs probabilities instead of text. It cannot hallucinate, runs faster than traditional LLMs, and charges per billion input tokens rather than per million, making it dramatically cheaper for software automation tasks. Early adopters report striking results: Vercel saw 5 to 18 times faster performance replacing ChatGPT, while another developer found Jev 10 to 20 times cheaper than Gemini for email classification. Use cases include agent monitoring, model routing, and workflow automation. Demand briefly crashed the company's API at launch.