Reliable model swapping for any local OpenAI/Anthropic compatible server - llama.cpp, vllm, etc
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).