TypeSafe AI releases new model Jev that produces probabilities instead of text
Diogo Almeida's TypeSafe AI launched a transformer-based model named Jev this week to address automation limitations in current large language models.
Model avoids text output issues
The new system generates probabilities rather than natural language tokens. This design prevents hallucinations because users define possible outputs before the model runs. Input costs are metered by the billion instead of the million.
High demand strains service capacity
Developers expressed strong interest in the product immediately after release. The company temporarily lost access to its API due to excessive traffic. TypeSafe describes the model as a cheaper and more robust tool for software automation.
Founder explains previous disappointment
Diogo Almeida left OpenAI two years ago because he felt existing models were not useful for automation. He argued that optimizing for human language does not help computers perform tasks. The new architecture aims to solve the problem of useless capabilities.
Reported by one outlet
Only one outlet has published this. Nothing here has been checked against a second report, so read it as that outlet's account and follow the link below for the original.
Reported by
1 independent outlet. Headline as published. Links open the original report.