[Rastonbury]

GPU VRAM is the bottleneck currently, check out r/localLlama for benchmarks and calculators for what models can fit into what cards approximately