▌ GitHub radar
Open Source Reimagines Sakana's LLM Router
An open reimplementation of Sakana AI's Fugu — the orchestrator that routes queries to the best model in your pool and measured 107% over any single model alone.
OpenFugu is an open reimplementation of Sakana AI's proprietary Fugu system — an orchestrator that uses a tiny 0.6B routing model to analyze each query and direct it to the most capable worker in a pool of LLMs. The measured result is 107% of the best single worker, meaning multi-model routing actually outperforms sticking with your single strongest model. The project covers all four stages: reading the architecture, running demos, training a custom router, and serving via API.
Why a vibe-coder should care
Instead of paying for one expensive model for every task, you could route simple requests to fast cheap models and only send hard questions to the heavyweights. OpenFugu makes that architecture buildable without proprietary access — you train your own router on your own data.
More finds