TypeSafe AI launches Jev, a decision model that skips text generation
San Francisco lab TypeSafe AI emerged from stealth on September 15, 2026 with $40M in seed funding and Jev, a model that returns typed answers with calibrated probabilities instead of generated text. Responses take 70-500 ms, input costs $0.042 per million tokens and output is free.
- Jev takes unstructured state plus typed questions and returns structured answers, never free text
- Input priced at $0.042 per million tokens with free output and 70-500 ms latency
- Post-training method RLCD rewards honest probabilities over pleasing prose
- Lab founded by Diogo Almeida, co-inventor of the RLHF method used for InstructGPT and ChatGPT
Read next
AI