Claude-Router (github.com)

The recent launch of Claude Router marks a significant advancement in intelligent model routing for Claude Code, automatically directing queries to the most suitable Claude model—Haiku, Sonnet, or Opus—based on task complexity. This innovation promises to cut costs by up to 98% for simpler queries without compromising quality, making it a game-changer in optimizing AI model usage. Unlike existing solutions that handle multi-provider routing, Claude Router focuses on intra-model optimization, streamlining the process and eliminating the need for manual model switching. Claude Router employs a UserPromptSubmit hook to classify queries almost instantaneously using a hybrid classification system. Simple queries can be routed to the cost-effective Haiku model while more intricate tasks are allocated to Sonnet or Opus. This strategy not only enhances response time—with Haiku delivering results 3-5 times faster than Opus—but also reduces GPU compute requirements and inference latency. As a result, users can enjoy extended usage limits and longer sessions while benefiting from more efficient resource allocation across Anthropic's infrastructure. Overall, Claude Router’s automatic and intelligent routing system enhances the user experience while promoting sustainable AI practices.