I Made a Mistake Installing vLLM on My Mac. My Disk Thanked Me for It.

2026-04-23 · Source: LLM on Medium · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, medium

Summary

A developer attempting to install vLLM on an M2 MacBook unexpectedly recovered 37 GB of disk space after encountering low storage issues. Initially, the installation failed due to platform incompatibility, as the commands were for Linux x86, not Apple Silicon. With only 2.2 GB free on a 228 GB disk, the developer investigated the `~/Library` folder, which contained 59 GB of data. Standard cleanup methods yielded only 3-4 GB. Deeper investigation revealed significant storage consumption by iOS simulators (18 GB), Claude desktop app VM bundles (9.2 GB), old Xcode archives (3.0 GB), and various application update caches. Targeted commands like `xcrun simctl delete unavailable` and manual deletion of specific directories led to the substantial recovery, increasing free space from 10 GB to 47 GB.

Key takeaway

For macOS developers struggling with disk space, your `~/Library` folder is likely a hidden culprit. You should prioritize checking and clearing developer-specific caches like old iOS simulators, Xcode archives, and large AI desktop app VM bundles. This targeted approach will yield far more significant storage recovery than generic cache clearing, ensuring you have ample space for new tools and projects.

Key insights

Developers can recover significant macOS disk space by targeting hidden, application-specific caches.

Principles

• Standard cleanup is often insufficient for developers.
• Application support folders hide large files.
• Simulators and old builds accumulate silently.

Method

Identify large developer-specific directories in `~/Library`, such as `CoreSimulator`, `Application Support`, and `Xcode/Archives`, then use targeted commands or manual deletion to clear outdated files.

In practice

• Run `xcrun simctl delete unavailable` for iOS simulators.
• Check `~/Library/Application Support/[App]/vm_bundles` for AI apps.
• Clear `~/Library/Developer/Xcode/Archives` for old builds.

Topics

• macOS Storage Optimization
• Developer Disk Usage
• vLLM Installation
• Xcode Simulator Management
• Claude Desktop App

Code references

• vllm-project/vllm

Best for: Software Engineer, Machine Learning Engineer

Related on AIssential

• I Deployed Local LLMs in Production for a Year. Part 1: The Mental Model — Production LLM deployment requires understanding hidden queues, KV cache memory, and multi-phase model loading beyond basic tutorials.
• Fine-Tune LLMs on Your Mac: DagsHub + Apple MLX-LM Integration — DagsHub and Apple MLX-LM enable full MLOps for local LLM fine-tuning on Apple Silicon.
• How to Cache vLLM Model in FastAPI for Faster Inference — Caching VLM/LLM models in FastAPI's global scope significantly reduces inference latency by avoiding repeated initialization.
• Semantic Caching for LLMs: TTLs, Confidence, and Cache Safety — Production-ready semantic caches require hardening beyond basic functionality to ensure safety and reliability.
• KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference — KVDrive optimizes long-context LLM inference by holistically managing KV cache across GPU, DRAM, and SSD tiers.
• v2.1.55 — The releases for Claude Code, spanning versions 2.1.69 down to 1.0.0, introduce significant advancements in AI agent capabilities, plugin extensibility, and user experience.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by LLM on Medium.