Workload

Local AI

Local AI is almost all GPU VRAM. 12 GB is the floor for useful image models; 24 GB+ for larger language models.