OpenRouter launches Jev Router to optimize AI model selection

OpenRouter has launched ‘Jev Router’, an automated model routing system powered by the Jev model released by TypeSafe AI. Unlike traditional routers that only analyze task types, Jev Router evaluates the specific difficulty and required precision of a prompt to assign the most appropriate AI model.

This approach aims to optimize efficiency by directing simple tasks to smaller, cheaper models while reserving high-end, expensive models for complex reasoning. In benchmarks involving four types of AI agent tasks, Jev Router successfully completed 82% more tasks than OpenRouter’s existing ‘Auto Router’. The system is designed for speed and operates under zero data retention terms, meaning prompts are not stored or used for training.

In related technical evaluations, developers have been testing various browser agent configurations using Jev. Comparisons between Claude Code with Playwright MCP and Jev-based agents like ‘jev-ultrafast’ and ‘laya-ultrafast’ show significant differences in execution speed and reliability for web automation tasks.