Aulafy
ESEN
Courses
Courses/Claude Code + Local AI/Minimum Hardware for Local AI in 2026
Lesson

Minimum Hardware for Local AI in 2026

The question isn't "which computer runs AI," but what experience you want: lightweight chat, RAG with documents, local coding, or agents. Each level demands a different mix of RAM, VRAM, and patience.

  • Choose hardware based on your real use case, not marketing.
  • Understand the difference between RAM, VRAM, and context.
  • Avoid buying the wrong hardware for Ollama, RAG, and local coding.

Quick reference by goal

TerminalCode
Aprender y probar:
  RAM: 16 GB
  VRAM: 6-8 GB o Apple Silicon con memoria unificada
  Modelos: 3B-8B Q4

RAG privado con documentos:
  RAM: 32 GB recomendado
  VRAM: 8-12 GB
  Modelos: 7B/8B Q4-Q5 + embeddings locales

Coding local razonable:
  RAM: 32 GB
  VRAM: 12-16 GB
  Modelos: Qwen/DeepSeek coder 7B-14B Q4-Q5

Agentes y tareas largas:
  RAM: 64 GB o más
  VRAM: 16-24 GB o más
  Modelos: 14B-32B, contexto controlado y logs

NVIDIA, AMD, and Apple Silicon

  • NVIDIA: usually the most straightforward path for GPU acceleration on Windows/Linux thanks to the CUDA ecosystem.
  • AMD: can work very well, but depends more on drivers, ROCm/Vulkan, operating system, and tooling.
  • Apple Silicon: unified memory helps a lot; look at total RAM and bandwidth, not just the chip name.
  • CPU-only: fine for learning and small models, but don't expect fast agents.

Checking your hardware

TerminalCode
# Windows / PowerShell
systeminfo | findstr /C:"Total Physical Memory"
wmic path win32_VideoController get name,adapterram
nvidia-smi

# macOS
system_profiler SPHardwareDataType
system_profiler SPDisplaysDataType

# Linux
free -h
lspci | grep -Ei "vga|3d|display"
nvidia-smi
Debug and protect your workOllama, vLLM, SGLang, and MLX: What to Use