RouteLLM GitHub repository (LMSYS LLM router)

README: 'RouteLLM is a framework for serving and evaluating LLM routers.' Core features as stated: drop-in replacement for OpenAI's client, or it can launch an OpenAI-compatible server, to route simpler queries to cheaper models; trained routers provided out of the box; claims routers 'reduce costs by up to 85% while maintaining 95% GPT-4 performance on widely-used benchmarks like MT Bench'; claims the same performance as commercial router offerings at '>40% cheaper'. Install via 'pip install routellm[serve,eval]'. Quickstart uses Controller with routers=['mf'], strong_model='gpt-4-1106-preview', weak_model='anyscale/mistralai/Mixtral-8x7B-Instruct-v0.1'. Linked blog lmsys.org/blog/2024-07-01-routellm/ and paper arXiv 2406.18665. IMPORTANT: all performance figures in the README are the authors' own and are VENDOR CLAIMS by this repo's standards, not independent measurements. API metadata 2026-10-04: created 2024-06-03, LAST PUSH 2024-08-10, 5568 stars, 441 forks, 44 open issues, not archived (so: not formally deprecated, but effectively dormant for about two years).