quantprobe
An open-source local LLM deployment and memory-placement planner that helps developers produce feasible quantization, tiering, and launch plans for running large models on consumer hardware.
Tool overview
Adoption verdict: quantprobe is worth trying if you repeatedly guess whether a local model will fit, download it anyway, and then hit VRAM or RAM limits. It is not a chat app, hosted inference service, or model training framework. A better analogy is a capacity planner and placement advisor for local LLM deployment: it helps estimate what is realistically runnable on your machine before you fully commit.
Based on the evidence, its practical role is to inspect hardware and estimate feasible quantization, VRAM/RAM/disk tiering, CPU/GPU splits, and related placement strategies, sometimes even before downloading a candidate model. The GitHub repo and author posts support the core claim about what it does, while one longer Japanese post gives a stronger workflow description than simple reposts do. There are also small test snippets, such as an estimate of about 0.3 tok/s for a specific quantized model on a GTX 1060 6GB setup, which is more useful than hype because it suggests conservative prediction rather than blanket performance claims.
On cost and difficulty, the evidence only supports that it is available as an open-source repository.