Saving money is the real deal; a 20% reduction in costs is too friendly to the agent team.

CryptoWorld News reports that Factory has announced the launch of a Model Router, a system capable of automatically selecting the most suitable AI model, reducing token consumption costs by 20% to 25%. The new router is specifically designed for agent workflows, where it intelligently assesses task complexity at startup, dispatching simple tasks to low-cost models and only invoking cutting-edge models like Claude Opus 4.7 for high-difficulty logical reasoning. Benchmark tests show that the new routing architecture achieved approximately a 99% pass rate of Claude Opus 4.7 on the Terminal-bench 2 test set, with a 20% reduction in per-session costs. On Legacy-bench, it achieved a performance equivalent to 96% of cutting-edge models while reducing costs by 25%. Factory