Hardware configuration.

Why this hardware?

Apple Silicon gives strong performance per watt for inference. The Mac Mini fits in an office or server closet, stays quiet, and avoids cloud lock-in. We size and recommend hardware based on your team size and model mix—contact us for a tailored setup.

About custom hardware.

The Mac Mini is our reference build, but your setup may need something different: more seats, heavier models, or existing servers. We can spec and guide custom hardware—bigger Apple Silicon machines, Linux workstations, or headless servers that run Ollama and your RAG stack.

• Larger Apple Silicon (Mac Studio, Mac Pro) for higher concurrency and bigger models.
• Framework Desktop—compact 4.5L AMD Ryzen AI Max system (up to 128GB LPDDR5x unified memory) with very strong iGPU throughput for local AI.
• Minisforum G7 series and Framework Laptop 16 (Max 395) are strong AMD Ryzen AI 300/Strix Halo options for Linux or Windows deployments.
• Linux workstations or servers with NVIDIA/AMD GPUs for maximum throughput.
• Existing on‑prem hardware—we help you size and install the Local AI stack on what you already have.

Framework Desktop note.

Framework Desktop is a compact desktop platform built around AMD Ryzen AI Max chips (including Ryzen AI Max+ 395 options) with high-bandwidth LPDDR5x unified memory configurations up to 128GB. For local AI, this is especially useful when you want strong integrated GPU performance without a large tower build. It is also a good fit if you prioritize modular, serviceable hardware and long-term repairability.