TypeSafe AI introduced the Jev model for structured automation

Company TypeSafe AI released Jev — a new frontier model specifically optimized for structured automation tasks. Developers claim that the system's architecture allows for a significant increase in performance compared to existing analogs.

Key performance indicators include a reduction in operation execution costs by 40–400 times. Data processing speed increases by 20–200 times at the same time. The model is designed to work in environments where result predictability and strict adherence to data formats are critical.

One of the main features of Jev is the guarantee of obtaining type-safe output. The system eliminates hallucinations due to strict binding to data schemas, making it suitable for integration into software pipelines that require typed responses without the need for additional validation or parsing.

The technology is aimed at developers creating complex agent systems and automated workflows. Using the model allows reducing delays when executing API requests and lowering infrastructure costs when scaling high-load applications.

The release is aimed at solving the problem of instability of output of large language models in corporate tasks. Implementing Jev allows replacing standard LLMs in scenarios where high accuracy of information structuring is required with minimal resource costs.