AI stack advisor

felch.ai — AI stack advisor

Will this model run on your box? Which scores are actually comparable? A catalog with honest license classes, a benchmark desk that refuses apples-to-oranges tables, and a hardware fit advisor that shows its work.

Hardware fit

VRAM, RAM, quant, context → a verdict band with the arithmetic exposed.

Benchmarks

Same-test scores only. Differing harnesses never share a table.

Model catalog

Filters for license class, size, modality, and local-runnability.