TypeSafe AI Launches Jev Model, Offering 5-18x Speed Improvement Over Luna

TypeSafe AI's new transformer-based model, Jev, has demonstrated a speed improvement of 5 to 18 times compared to OpenAI's ChatGPT Luna, as reported by Vercel, which has transitioned to using Jev. Founded by Diogo Almeida, the inventor of reinforcement learning from human feedback (RLHF), TypeSafe AI aims to provide more efficient AI solutions. Jev also offers a cost advantage, being 10 to 20 times cheaper than Gemini, another model tested by Bryo AI for classifying business emails. Almeida, who left OpenAI two years ago to establish TypeSafe AI, stated, "We have lightning in a bottle, and yet it is not useful," emphasizing the potential of Jev in automating workflows. Nikhil Mudholkar, CTO of Bryo AI, noted that Jev's ability to return real probabilities makes it particularly suitable for workflow automation. The model has shown output confidence scores ranging from 50% to 95%, prompting users to assess the reliability of its predictions, as highlighted by Armin Ronacher, CTO of Earendil. The launch of Jev coincides with an event in San Francisco from October 13 to 15, further showcasing TypeSafe AI's advancements in the AI landscape.