Buying a Mac for AI: The Only Guide You Need

Which Mac should you buy for local AI in 2026, and is the Mac you already own enough? A buying guide by model size, speed and memory, with a free tool that answers it for your own machine. We take five open models, from Qwen 3.8 27B on a 32 GB Mac mini to GLM-5.3 on a 512 GB Mac Studio, and put them against the new M5 and M6 Macs, the M5 Ultra Studio, NVIDIA's DGX Spark and an M1 Max MacBook Pro from 2021. What each model can actually do for you, how fast it writes on each machine, what fits and what doesn't, what it costs, and where a used Mac is the smarter buy. Then the tool that does all of this for your own Mac: LLM Sizer, free, no sign-up. 🛠 The free tool: https://theaibridges.com/tools/llm-si... 📐 How the numbers are made: https://theaibridges.com/tools/llm-si... 📬 The newsletter (AI for business, no hype): https://theaibridges.substack.com 100,000 views on this video and I open-source the tool's code. Chapters: 00:00 Should you buy a new Mac for AI? 01:21 What do you want AI to do? Five model sizes, from a 32 GB mini to a 512 GB Studio 08:17 How fast will AI run on your Mac? The six things that set the speed 15:04 New Macs vs the models: what fits, what doesn't, and how fast 23:45 Why not just use the cloud models? 25:52 The caveats of running AI locally 28:47 So which Mac should you buy? 31:40 The free tool, page by page: LLM Sizer 38:00 What we covered Every speed in this video is an estimate from the tool's published formula, calibrated on measured runs; the M5 Ultra numbers are assumptions until someone measures one. Prices are Apple and NVIDIA list prices as of 2 September 2026. Connect: YouTube:    / @alexgrankin   LinkedIn:   / alex-grankin   X: https://x.com/alex_grankin Site: https://theaibridges.com #LocalAI #MacStudio #LLM