LM Studio Australia: Run Private AI on Your Laptop

Why LM Studio Is the Best Starting Point for Local AI

If you have read our complete guide to running AI offline at work, you know there are several tools for running AI locally: Ollama, Jan.ai, GPT4All, and LM Studio. Each has its place.

But if you have never touched a command line and just want something that works like ChatGPT, but runs entirely on your laptop, LM Studio is where you should start.

I have set up local AI for dozens of corporate clients. The pattern is always the same: non-technical users try Ollama, get stuck at the terminal, and give up. Then I show them LM Studio, and within 10 minutes they are chatting with a local AI model.

No coding. No command line. Just a clean interface, a download button, and a chat window.

What Makes LM Studio Different

LM Studio is not just a wrapper around command-line tools. It is a purpose-built desktop application with a proper user interface. Search for models, download them, and start chatting, all from one window. It is also completely free, with no hidden subscriptions or premium features.

What You Will Learn in This Guide

This guide covers everything from first download to advanced features:

By the end, you will have a fully functional private AI assistant on your laptop that never sends data to the cloud.

System Requirements: Can Your Computer Run This?

Before downloading, let us make sure your hardware can handle local AI. Here is the honest truth about what you need.

macOS Users

If you have an Apple Silicon Mac (M1, M2, M3, or M4), you are in a great position. The unified memory architecture means your Mac can run surprisingly large models. An M1 MacBook Air with 16GB can comfortably run 7-8B parameter models.

Requirements:

• macOS 13.4 (Ventura) or later
• M1/M2/M3/M4 chip (Intel Macs work but are slower)
• 16GB RAM recommended

Windows Users

Most business laptops from the past 3-4 years will work fine for smaller models.

Requirements:

• Windows 10 or 11
• CPU with AVX2 support (almost all CPUs since 2013)
• 16GB RAM recommended
• NVIDIA GPU helps but is not required

Linux Users

LM Studio ships as an AppImage, so it works on most distributions without installation headaches.

Requirements:

• Ubuntu 20.04 or equivalent
• x64 or ARM64 architecture
• Same RAM and storage recommendations as Windows

Step-by-Step: Downloading and Installing LM Studio

This takes about 2 minutes.

Step 1: Go to lmstudio.ai

Open your browser and navigate to lmstudio.ai. The site automatically detects your operating system and shows the appropriate download button.

Step 2: Download the Installer

Click the download button. The file is about 150-200MB depending on your platform:

• macOS: LM-Studio-x.x.x-arm64.dmg (Apple Silicon) or LM-Studio-x.x.x-x64.dmg (Intel)
• Windows: LM-Studio-x.x.x-x64.exe
• Linux: LM-Studio-x.x.x.AppImage

Step 3: Install

macOS:

1. Open the downloaded .dmg file
2. Drag LM Studio to your Applications folder
3. Right-click the app and select "Open" (first time only, to bypass Gatekeeper)

Windows:

1. Run the downloaded .exe file
2. Click "Yes" on the UAC prompt
3. Follow the installer (Next → Next → Finish)

Linux:

1. Make the AppImage executable: chmod +x LM-Studio-x.x.x.AppImage
2. Double-click to run, or use: ./LM-Studio-x.x.x.AppImage

Step 4: First Launch

When you first open LM Studio, you will see a welcome screen. The app may check for updates, let it update if prompted. Version 0.3.36 (as of January 2026) includes important bug fixes and new model support.

The LM Studio Interface: A Complete Tour

LM Studio has a clean, modern interface. Once you understand where things are, everything clicks into place.

The Left Sidebar

The sidebar contains five main tabs:

The Discover Tab (Model Search)

This is your "app store" for AI models. When you click the magnifying glass icon, you see a search bar at the top and curated model suggestions below.

What you will see:

• Search bar: Type model names like "llama", "mistral", or "qwen"
• Trending models: Popular downloads this week
• Model cards: Each model shows size, download count, and compatibility info

The Chat Tab

This is where you spend most of your time. It looks and feels like ChatGPT:

• Model selector (top): Choose which model to load
• Conversation list (left): Your chat history, organised in folders
• Chat area (centre): The conversation itself
• Input box (bottom): Where you type messages
• Settings panel (right, collapsible): Fine-tune model behaviour

The My Models Tab

Lists all models you have downloaded. From here you can:

• See file sizes and locations
• Delete models you no longer need
• Check compatibility with your hardware

Downloading Your First Model: The Right Choice Matters

This is where many beginners get stuck. There are thousands of models available, and picking the wrong one means either poor performance or failed loads.

My Recommendation: Start with Llama 3.2 8B

If you have 16GB of RAM, download Llama 3.2 8B Instruct. Here is why:

• Well-balanced size (about 5GB download)
• Great at general tasks: email drafting, summarisation, Q&A
• Runs smoothly on most modern laptops
• Large community, lots of support online

How to Download a Model

1. Click the Discover tab (magnifying glass in sidebar)
2. Search for "llama 3.2" in the search bar
3. Look for "llama-3.2-8b-instruct" in the results
4. Check the file size - Look for a quantised version (Q4_K_M or Q5_K_M)
5. Click Download on the version that fits your hardware

Understanding Model Names:

Model filenames contain important information:

llama-3.2-8b-instruct-q4_k_m.gguf
│       │    │        │
│       │    │        └── Quantisation level (smaller file, slightly less accurate)
│       │    └── Fine-tuned for instruction following (conversations)
│       └── 8 billion parameters
└── Model family (Meta's Llama)

Quantisation quick guide:

• Q4_K_M: ~4GB file, good balance (recommended for most users)
• Q5_K_M: ~5GB file, slightly better quality
• Q8_0: ~8GB file, highest quality quantised version
• F16: ~16GB file, full precision (only for high-end hardware)

Download Progress

Downloads can take 5-30 minutes depending on your internet speed and model size. The progress bar shows at the bottom of the screen. You can queue multiple downloads.

Your First Conversation: From Zero to Chatting

Once your model is downloaded, you are one click away from using it.

Step 1: Go to the Chat Tab

Click the chat bubble icon in the left sidebar. You will see an empty conversation area with a "Load a model to start chatting" prompt at the top.

Step 2: Load Your Model

1. Click the model selector dropdown at the top of the chat area
2. Select the model you downloaded (e.g., "llama-3.2-8b-instruct-q4_k_m")
3. Wait for it to load (10-60 seconds depending on hardware)

What happens when loading:

• The model file is read from disk
• Weights are loaded into RAM (or VRAM if using GPU)
• You will see memory usage increase in your system monitor
• A green indicator appears when ready

Step 3: Start Chatting

Type your first message in the input box at the bottom and press Enter. Try something like:

Summarise the key benefits of working from home in 3 bullet points.

The AI will generate a response, streaming word by word. On a MacBook M1 with 16GB RAM, expect 10-20 tokens per second with an 8B model, fast enough to feel interactive.

Step 4: Continue the Conversation

Unlike single-prompt tools, LM Studio maintains conversation context. The AI remembers what you discussed earlier in the chat. Ask follow-up questions, request revisions, or change the topic.

Understanding the Settings Panel

Click the gear icon (or the collapsible panel on the right) to access model settings. These affect how the AI responds.

Key Settings Explained

Temperature in practice:

• 0.0-0.3: Factual, consistent responses (good for data extraction, coding)
• 0.5-0.7: Balanced (good for general conversation, emails)
• 0.8-1.0+: Creative, varied responses (good for brainstorming, creative writing)

System Prompts

The system prompt tells the AI how to behave. By default, most models use something like "You are a helpful assistant."

For work use, consider custom system prompts like:

You are a professional business assistant. Respond in a formal tone
suitable for corporate communication. Be concise and action-oriented.
Use Australian English spelling.

Uploading Documents for Context

One of LM Studio's most useful features is document upload. You can attach files and ask the AI questions about them.

Supported File Types

• PDF - Reports, contracts, manuals
• DOCX - Word documents
• TXT - Plain text files

How to Use Document Upload

1. Click the attachment icon (paperclip) in the chat input area
2. Select your file
3. Wait for processing (may take a moment for large PDFs)
4. Ask questions about the document

Example prompts after upload:

Summarise the key points of this document in 5 bullet points.

What are the payment terms mentioned in this contract?

List all the action items from these meeting notes.

Limitations to Know

• Context length: Long documents may be truncated to fit the model's context window
• Accuracy: Local models are less accurate than GPT-4 for complex document analysis
• Formatting: Tables and complex layouts may not parse perfectly

The Local Server: For Developers and Integrations

LM Studio can run as a local API server, compatible with OpenAI's API format. This means any application designed for OpenAI can work with your local model instead.

Why Use the Local Server?

• Connect other apps: Use LM Studio with VS Code extensions, note-taking apps, or custom scripts
• No code changes: Apps that work with OpenAI just need a different base URL
• Free unlimited usage: No API costs, no rate limits

Starting the Server

1. Click the Server tab (server icon in sidebar)
2. Select a model to serve
3. Click "Start Server"
4. Note the URL (usually http://localhost:1234/v1)

Connecting Applications

Point applications to:

• Base URL: http://localhost:1234/v1
• API Key: Any string (e.g., "lm-studio"), not validated

Most OpenAI-compatible tools just need these two settings changed.

Troubleshooting Common Issues

After helping many first-time users, these are the problems I see most often.

"Model failed to load" or Crashes on Load

Cause: Not enough RAM for the selected model.

Fix:

1. Close other applications to free memory
2. Try a smaller model (7B instead of 13B, or more quantised version)
3. Reduce context length in settings
4. On NVIDIA: reduce GPU layers to offload less to VRAM

Very Slow Responses (< 1 token/second)

Cause: Model running on CPU when GPU would be faster, or insufficient resources.

Fix:

1. Check GPU layers setting, increase if you have VRAM available
2. Use a more quantised model (Q4_K_M instead of Q8_0)
3. Close background applications
4. On Mac: ensure Metal acceleration is enabled (it is by default)

"Download failed" or Stuck Downloads

Cause: Network issues or Hugging Face server problems.

Fix:

1. Click retry on the download
2. Try a different quantisation of the same model
3. Check your internet connection
4. Try again in an hour (server may be busy)

Model Outputs Gibberish or Repetitive Text

Cause: Incorrect chat template or model issue.

Fix:

1. Try a different model (some models are poorly optimised)
2. Adjust temperature (lower it to 0.5-0.7)
3. Use an "instruct" version if available (trained for conversations)

Best Practices for Corporate Use

If you are using LM Studio at work, here are the talking points and best practices for corporate environments.

How to Explain LM Studio to IT

When asking IT for approval, focus on the security story:

1. No network traffic: After downloading the model, LM Studio works entirely offline
2. No data collection: Your conversations stay on your hard drive
3. No account required: No login, no usage tracking, no telemetry
4. Open source models: Community-vetted, transparent weights

What to Use It For (and What to Avoid)

Good use cases:

• Drafting emails and documents
• Summarising meeting notes
• Explaining technical concepts
• Cleaning up rough notes
• Code explanation and debugging help

Use cloud AI instead for:

• Very long document analysis (100+ pages)
• Tasks requiring the latest information
• Complex multi-step reasoning
• When accuracy is critical (verify important outputs)

Where to Go From Here

You now have a working local AI assistant. Here are the next steps to get more value from it.

1. Experiment with Different Models

Once comfortable with Llama 3.2, try:

• Mistral 7B: Faster, great for quick tasks
• DeepSeek-R1 7B: Excellent for coding and reasoning
• Qwen 2.5: Strong multilingual support

2. Create Custom Presets

Save your favourite system prompts and settings as presets. Create different presets for:

• Email drafting (formal tone, concise)
• Brainstorming (high temperature, creative)
• Technical work (low temperature, accurate)

3. Explore the Server Mode

If you use coding tools like VS Code with AI extensions, point them to LM Studio's local server for privacy-focused code assistance.

4. Read the Companion Guides

• How to Use AI Offline at Work: Complete Guide, The broader context on local AI and why it matters
• OpenAI vs Claude vs Ollama: The Definitive Guide, Compare local options with cloud alternatives

Wrapping Up

LM Studio removes the biggest barrier to local AI: complexity. You do not need to understand Python, Docker, or command-line tools. You just need a decent laptop and 10 minutes.

Is it as capable as GPT-4? No. But for routine work tasks (email drafting, document summarisation, meeting notes, quick questions), it is more than good enough. And your data never leaves your machine.

That trade-off is worth it for many professionals, especially those in regulated industries or companies that block cloud AI tools.

Download LM Studio, grab a Llama 3.2 model, and give it a try. The worst case is you learn something new. The best case is you gain a private AI assistant that costs nothing to use.

Need Help Implementing Local AI Across Your Organisation?

LM Studio is perfect for individual use, but rolling out local AI across a team of 10, 50, or 100+ employees requires proper planning. IT policy alignment, model governance, hardware requirements, and compliance documentation do not configure themselves.

Solve8 helps Australian businesses deploy private AI infrastructure that meets enterprise security requirements while keeping data within Australian borders.

What we offer:

• Free AI Assessment, Understand your privacy requirements and best-fit solutions
• Local AI Strategy, Model selection, hardware specs, and deployment planning
• Implementation Support, We configure, deploy, and train your team
• Compliance Documentation, Privacy Act alignment and IT policy templates

Book a free 30-minute consultation →

No sales pitch. Just honest advice on whether local AI makes sense for your organisation.

Related Resources:

• Offline AI for Corporate Use: Complete Guide
• Data Sovereignty Guide for Australian Business
• Private AI Infrastructure Services
• AI Strategy & Roadmap Services

Sources:

• LM Studio Official Website
• LM Studio Documentation
• LM Studio 0.3.36 Release Notes

Solve8 is an Australian AI consultancy helping businesses navigate the complex landscape of AI implementation. Based in Brisbane, serving clients across Australia. ABN: 84 615 983 732