According to Visionary Beating monitoring, the software development AI platform Factory announced the launch of the Factory Router model router. The system is able to automatically select the most suitable AI model for different software development tasks, maintaining state-of-the-art large model performance while reducing token consumption costs by 20% to 25%.
The new router is designed specifically for agent workflows. The system intelligently assesses complexity at task initiation, assigning simple refactoring or documentation writing to low-cost models, and only invoking cloud-edge models like Claude Opus 4.7 when encountering high-difficulty logic reasoning. Additionally, the model scheduling mechanism supports multi-service provider disaster recovery and maintains dialogue coherence through context compression technology when switching endpoints.
In benchmarks released by Factory, the new routing architecture achieved approximately 99% throughput on the Terminal-Bench 2 test set with Claude Opus 4.7 and reduced single-session costs by 20%. The system achieved a performance equivalent to 96% of state-of-the-art models on Legacy-Bench while decreasing costs by 25%. The Factory Router is currently available to a subset of testing users in the platform's command-line interface (CLI) and desktop application.
