ChatGPT Co-Inventor Launches Jev, A Frontier Model That Never Generates Text
0:00 / 0:50
News
ChatGPT Co-Inventor Launches Jev, A Frontier Model That Never Generates Text
calendar_today Date:
schedule Duration: 0:50
database
Summary Report
Jev from Typesafe AI is a System One model that outputs only typed decisions with calibrated confidence, claiming up to 200x the speed and 400x lower cost of frontier LLMs for the smart if-statements inside software.
- 01. No text generation: typed structured outputs with calibrated probabilities, trained with RLCD rather than RLHF
- 02. 70-500ms per answer, $0.042 per million input tokens, free output tokens; speed claims from Typesafe's own evals; waitlist only
Diogo Almeida, who co-created RLHF and ChatGPT at OpenAI, has launched Jev from his company Typesafe AI after two years in stealth. It's a System One model: no text generation, only typed structured decisions with calibrated probabilities, trained with a new method called RLCD. Aimed at classification, routing, scoring and extraction, it claims 70 to 500 milliseconds per answer, $0.042 per million input tokens, free output tokens and up to 200x the speed of frontier LLMs on Typesafe's own workflow evals. Waitlist only.
