I'm not the person you're replying to, but I do have a 64GB machine that I'd been planning to bump up to 128 right around the time the prices went through the roof. My uses are:
- VMs, I'm leaning on them more and more for sandboxing stuff I'm working on, both because of the rise in software supply chain threats, and to put guardrails around AI agents.
- Local LLMs experimentation, even pretty big MoE models (GPT OSS 120b) run pretty usably (~10 tokens/sec) with the latest tooling on a 16GB GPU and a lot of system memory.
- Even compared to a fast NvME drive, it's super nice to load a big dataset into memory and just process it right there, compared to working off of the disk.
Yeah, I have a 64GB M1 Max and can run local models pretty well. I bought it on release and even now it never feels slow. I may upgrade just because I want to move to the 14” since I travel more now.