Don’t Buy a Used Mac Studio for Local AI (Buy This Instead)
Used Mac Studio vs new AMD mini PC for local AI: which 128 GB machine is the better buy? Real prices, speed on gpt-oss-120b, and the software setup. On October 12, 2026, a used M1 Ultra Mac Studio with 128 GB of memory that you could actually order cost between $4,000 and $5,000, while a new AMD Ryzen AI Max+ 395 mini PC with the same memory and 2 TB of storage started at $3,299. On the standard small-model test the old Ultra writes roughly twice as fast; on gpt-oss-120b that lead shrinks to about a sixth, and prompt reading is basically a tie. For macOS buyers, a new M5 Max reads prompts about 2.7 times faster than a used 64-core M1 Ultra on that small-model test, but it costs $5,399. GPU memory differs by system: Linux gives the AMD box's GPU about 120 GB, Windows reserves 96 GB, and macOS holds back a portion by default until one terminal command raises the limit. Ollama's support page lists the 395 under ROCm on Linux, Windows can still run it on the GPU through Vulkan, and Macs get Apple's MLX framework. If you don't need macOS, the new AMD mini PC is the better value; if you do, a used M1 Ultra at about $4,000 suits you if you mostly wait on long generated answers, and the new M5 Max is the pick for long prompts or Apple's warranty. For builders and AI tinkerers weighing hardware for local model inference. Chapters: 0:00 Intro 1:33 The Prices 3:39 The Speed 6:45 The New Mac 8:15 The Software 10:08 Verdict Tools & resources mentioned: - llama.cpp: https://github.com/ggml-org/llama.cpp - Ollama: https://ollama.com - MLX: https://github.com/ml-explore/mlx - AMD Ryzen AI Max+ 395: https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html About The Stack The Stack helps you build with AI. Each video takes one tool, model, or workflow and shows how it works in a few focused minutes, with the real benchmarks and real costs. We go deep on Claude Code and Cursor for AI coding, AI agents and MCP servers, the open-source AI tools and GitHub repos most people miss, RAG and vector search, fine-tuning, and running local LLMs on your own machine with Ollama and LM Studio. We compare models like ChatGPT and Claude, test AI automation with Zapier, Make, and n8n, and flag the tools that actually ship. Subscribe for new breakdowns: https://www.youtube.com/@the-stack-ai?sub_confirmation=1 #MacStudio #MacMiniVsMacStudio #LocalAI
