A new web tool lets users input their GPU/Mac specifications, desired context length, and KV cache type to see which of about 3,000 popular GGUF‑quantized LLMs will fit in memory, including the exact llama.cpp launch command. It computes memory usage from each model’s header (layer count, KV heads, sliding window, etc.) and ranks the models accordingly. The tool is hosted on Hugging Face Spaces at LocalLLaMA/local-model-explorer.