Will it fit, and will it crawl?
FitLLM is for machines with RAM and a little disk, not an AI GPU. Think an Intel Core i5-8350U-class laptop with 32 GB RAM and about 32 GB free. RAM lets you load large models; a 15 W CPU makes them slow. Disk remains the hard limit.
Catalog that respects a 15 W CPU
Chat GGUFs, diffusion checkpoints, and image-to-mesh weights. Disk is the filter that matters.
Runtimes worth installing
Use GGUF engines for chat, stable-diffusion.cpp for pictures, and a small dedicated checker for meshes.
Playbook for a weak CPU
Use one quant per model family, short context, AVX2 builds, and realistic expectations.
Storage diet
One daily driver beats five quants of the same model. Keep cold storage on a USB SSD.
Phones and tablets
Modern phone silicon can beat an older laptop on small models, but RAM and thermals are the ceiling.
Saved picks
Your favorites and hardware profile are saved locally and to signed-in Popchat storage when available.