Pradhyuman Yadav

A new kind of AI model from a ChatGPT inventor is thrilling developers

By Pradhyuman,

Former OpenAI researcher Diogo Almeida founded TypeSafe AI to build models for software automation. The startup released Jev, a transformer-based model that produces probabilities instead of text. Users define outputs in advance, so the model does not hallucinate. TypeSafe AI meters input tokens by the billion and provides output tokens for free.

Software developers use Jev for task automation. Vercel software engineer Pranit Sharma reported that replacing OpenAI's ChatGPT Luna 5.6 with Jev for a safety classifier yielded results 5 to 18 times faster with higher accuracy. Bryo AI CTO Nikhil Mudholkar tested Jev against Gemini for email classification. He stated that Gemini was slightly more accurate, but 10 to 20 times more expensive.

Almeida stated that TypeSafe AI trained Jev on synthetic data using reinforcement learning from calibrated decisions.

Maintained by Pradhyuman.

Filed under: OpenAI, Google, Models & Research, Products & Agents

Related articles