It doesn't matter what 'most users' are doing. On a 64 GB Apple silicon Mac mini you can natively host mid sized and some larger quantised local models .. usin

It doesn't matter what 'most users' are doing. On a 64 GB Apple silicon Mac mini you can natively host mid sized and some larger quantised local models .. using Ollama. For example: Qwen3-Coder (32B), GLM-4.7 (or GLM-4 Variants), Devstral-24B / Mistral Large (Quantized)