Check if your computer can run private AI

Private AI runs on your own computer, so nothing you type leaves your desk. But it only works well if your computer has enough power. This guide shows you how to find the three numbers that matter, what they mean in plain English, and whether your computer is ready to try it.

Before you start

• The computer you want to use, switched on.
• About 10 minutes.
• A notepad to write down three numbers.

Want to do it yourself?

Difficulty

4 out of 10

Estimated completion time

20–40 minutes

Skills you’ll need: Find system memory, processor, and free disk space, then compare them with an app’s published requirements.

I’ll do it myselfI’d rather get help

How this helps your brand

Checking your computer first helps you avoid a private-AI setup that is too slow for daily work. Matching the task to available memory supports a practical workflow without promising that every device or model will perform alike.

1. 1

   Step 1: Find out how much memory your computer has

   Memory, also called RAM, is the computer’s short-term workspace. It’s the single biggest factor in what private AI you can run, because the whole AI model has to fit in it while it works. It’s not the same as storage space for your files.

   To find it:

   • On a Mac: open the Apple menu in the top-left corner and choose About This Mac. Look for “Memory”.
   • On Windows: open Settings, then System, then About. Look for “Installed RAM”.

   Tip: Write the number down in gigabytes (GB), for example 8 GB, 16 GB or 32 GB.

   Illustrated exampleFind the RAM number, not the storage number

   ● ● ● Mac example

   Apple menu → About This MacMacBook Air

   Chip

   Apple M2

   Memory

   16 GB 1

   ▦ Windows example

   Settings → System → AboutDevice specifications

   Processor

   Intel Core i5

   Installed RAM

   16 GB 1

   1Write down 16 GB of RAM. “Memory” and “Installed RAM” mean the same thing here.

   120 GB free on your disk?That is storage for files. It does not mean you have 120 GB of RAM.

2. 2

   Step 2: Note your chip and graphics

   On the same screen, note what processor (chip) your computer has. This tells you which private AI apps will run on it.

   • Mac: if it says M1, M2, M3, M4 or newer, you have Apple Silicon. These Macs share memory with the graphics chip, which suits private AI well. Older Intel Macs aren’t supported by some popular apps, such as LM Studio, so check each app’s requirements page.
   • Windows: note the processor name. If you also have a separate graphics card (often from NVIDIA or AMD), note its name and how much memory it has. You can usually find this in Task Manager under Performance. A graphics card with its own memory can make answers much faster, but it isn’t required to start.
3. 3

   Step 3: Check your free disk space

   Each AI model is a large file, often several gigabytes. You’ll want room for the app, at least one or two models, and space to spare.

   • On a Mac: open System Settings, then General, then Storage.
   • On Windows: open Settings, then System, then Storage.

   Tip: If you have less than about 20 GB free, clear some space before you install anything. That’s a rough guide; each model’s download page shows its exact size.

4. 4

   Step 4: Match your memory to a model size

   Models come in sizes, usually shown as a number followed by “B”, like 3B or 8B. The B stands for billions of parameters, which is a rough measure of how big and capable the model is. Bigger models give better answers but need more memory and run slower.

   Use these as rough starting points, not rules. Requirements change as new models and apps come out:

   • 8 GB: small models only. Fine for short drafts, rewording and quick summaries. Expect it to be slower, and close other apps while you use it.
   • 16 GB: comfortable for small and mid-size models. A good level for everyday writing and summarising.
   • 32 GB or more: room for larger, more capable models and longer documents.

   Tip: Before you download any model, check its page in the app or on its official site for the memory it needs. Apps like LM Studio often show whether a model is likely to fit your computer.

   Illustrated exampleMore memory gives your model more working room

   Your RAMRough starting point

   1. 8 GB
      Start small

      Short drafts, rewording and quick summaries.

      Close other apps; expect slower answers.
   2. 16 GB
      Everyday writing

      Small and mid-size models for writing and summarising.

      A comfortable starting point for everyday tasks.
   3. 32 GB+
      More room

      Larger models and longer documents.

      Check the particular model before downloading.

   3B / 8B

   Model size is a clue, not a memory requirement.“B” counts billions of parameters. Check the model’s stated RAM requirement against your computer’s memory.

5. 5

   Step 5: Check the app’s official requirements

   Pick the app you’re most likely to try and read its official system requirements page. The common beginner choices are LM Studio (lmstudio.ai), Ollama (ollama.com) and AnythingLLM (anythingllm.com). Each lists the operating systems and chips it supports.

   If your computer meets the requirements and you have at least 8 GB of memory, it’s worth trying. If not, a cloud AI tool with good privacy settings is the better choice for now, and you can come back to private AI when you next replace your computer.

I run a small business and I want to try private AI that runs on my own computer. I’m not technical, so please explain in plain English.

My computer: [for example, MacBook Air M2, or Dell laptop with Windows 11]
Memory (RAM): [for example, 16 GB]
Graphics card, if any: [name and its memory, or “none”]
Free disk space: [for example, 120 GB]
What I want to use it for: [for example, drafting emails and summarising documents]

Please:
1. Tell me whether my computer is likely to run private AI well, and what size of model to start with.
2. Tell me what speed to realistically expect for a short email draft.
3. Suggest which beginner app to try first, such as LM Studio, Ollama or AnythingLLM, and why.
4. List anything I should double-check on the official app and model pages, because requirements change. Don’t guess at exact numbers you aren’t sure of.