BeginnerMac / Apple 7 min read

M1 / M2 Mac 8GB: Realistic Ollama Limits

Unified memory is shared with macOS — here is what actually works on base MacBooks without swapping.

Verified stack

M1/M2 8GB · Ollama Metal · 3B Q4 or 7B Q2/Q3 · ctx ≤2K · close browsers

Last verified 2026-07-22

M1M28GB RAMOllamaMetal

Memory budget

macOS + apps use 3–4GB. That leaves ~4GB for the model on an 8GB Mac. Stick to 3B Q4 or 7B Q2/Q3 with short context. Close browsers before loading 7B.

text

M1 8GB safe picks:
  llama3.2:3b       → smooth chat
  qwen2.5:3b        → good Chinese
  phi3.5:mini       → fast responses

Avoid on 8GB:
  llama3.1:8b @ Q4  → swap thrashing
  any 14B+ model

Ollama settings

Set OLLAMA_NUM_PARALLEL=1 and keep context at 2048 for 8GB machines. Monitor Memory Pressure in Activity Monitor.

bash

export OLLAMA_NUM_PARALLEL=1
export OLLAMA_MAX_LOADED_MODELS=1
ollama pull llama3.2:3b
ollama run llama3.2:3b

Related guides

BeginnerMac / Apple6 min read

Mac M3 Max: The Ultimate Local LLM Setup

Maximise your Apple Silicon with Ollama. Run multiple models, set up an OpenAI-compatible API, and tune Metal GPU layers.

BeginnerMac / Apple6 min read

Mac M3 Pro: Realistic Model Limits

What actually fits in 18GB or 36GB unified memory with Ollama and llama.cpp.

BeginnerDocker5 min read

Docker Compose LLM Stack: Ollama + Open WebUI

A production-ready Docker Compose stack that gives you a local ChatGPT experience with one command.

Deployment guides are educational. Each model is subject to its own license — read the official Hugging Face model card before downloading or deploying.