The author released an open‑source script, whatmodel.sh, that takes a user's RAM and GPU specs and returns a ranked list of AI models that can run locally, including the command to launch each model. Each model receives a unified score derived from 19,731 published benchmark results, with links to the original sources, allowing comparison against API‑based models. The tool shows that even a 32 GB Mac can now run a model scoring on par with GPT‑5, such as Qwen3.8 27B at 4‑bit quantization, entirely for free.