What are the RAM requirements for running AI models locally on a Mac?

Running AI models locally, even smaller ones, often necessitates a minimum of 32GB of RAM for adequate performance.

Can an older MacBook Pro with 16GB RAM run large language models?

An older MacBook Pro with 16GB of RAM can struggle significantly, leading to extreme slowness and lengthy processing times for AI models.

Why is there a growing demand for RAM in AI?

DRAM is becoming increasingly memory-intensive as cloud data centers consume more of it to run large language models.

Home / Technology / Old Mac Struggles with New AI Demands

Old Mac Struggles with New AI Demands

1 Feb

•

Summary

Running large language models locally requires significant RAM, ideally 32GB.
Older machines with 16GB RAM can experience extreme slowness with AI models.
Newer AI models demand more powerful hardware, often exceeding older specs.

An attempt to run open-source large language models (LLMs) locally on a three-year-old MacBook Pro with 16GB of RAM revealed significant hardware limitations. While the machine handles everyday tasks, it faced extreme performance issues, including lengthy processing times, when attempting to run AI models.

Models like glm-4.7-flash, weighing 19 gigabytes, took over an hour to generate a simple response. Even a purportedly faster model, gpt-oss:20b, exhibited slow performance. These experiences underscore that running LLMs effectively often requires at least 32GB of RAM, a specification now considered a minimum for information workers engaging with AI.

The investigation indicated that memory-intensive AI operations are pushing the boundaries of even relatively recent hardware. The rising cost and demand for DRAM by cloud data centers further complicate accessibility for individuals seeking to run AI models on personal machines. This situation suggests a potential need for hardware upgrades to keep pace with AI advancements.