How much memory do you need to run AI on a laptop?

Short answer: The model has to fit in your laptop's memory, with room left over for the document and your other apps. More memory lets you run bigger models, and bigger models do harder jobs better. The AI app LM Studio recommends 16 GB or more of memory. Our four tiers, A to D, show what each level of memory can do, using measured results rather than guesses.

Where the model lives#

A model needs fast memory while it works. Where that memory sits depends on the laptop.

• Laptops with a separate graphics card have two pools: the main memory (RAM) and the card's own memory (VRAM). AI runs fastest when the whole model fits in the card's memory.
• Macs with Apple chips use one shared pool, called unified memory. Apple says the whole chip can reach a single large pool of memory. That helps it run larger AI models on the device. See Apple's note on the M5 chip.

If a model does not fit in graphics memory, Ollama can use the main memory instead. Ollama's quickstart says responses may then be slower. Ollama's ollama ps command shows where a model is running: all on the graphics chip, all on the processor, or split. See the Ollama quickstart and FAQ. Our scorecards report this as "fits in graphics memory".

Three things use memory#

1. The model. A good first guide is the size of the model file. Ollama's quickstart uses one example model with a download of about 7.2 GB, and recommends 8 GB of available graphics memory or unified memory on a Mac.
2. The document you give it. Longer inputs need more memory. Ollama says that a larger context length increases the memory a model needs. See Ollama's context length page. In our tests, every laptop gets the same context size, so results are fair. See our method.
3. Everything else. Your operating system and open apps share the same memory. This is why a laptop with 16 GB cannot give the model all 16 GB. We measure peak memory on every test.

Model sizes and compression#

Model sizes are written in billions of parameters, such as 7B. A bigger number usually means a more capable model and a larger file.

Compression makes a model file smaller. LM Studio describes it as shrinking the file while giving up some quality. We test at one common compression setting, called Q4_K_M, so every laptop faces the same choice. See LM Studio's download guide.

Our four tiers#

We sort laptops into tiers by the memory the AI can use. Each tier page lists the laptops we have tested in it and what they can do.

These tier definitions come from our test protocol. See how we test.

What each tier can actually do is measured. It is not guessed. Open a tier page to see the results, then look at the jobs you care about, for example contract summaries, handbook Q&A and customer replies.

What the app makers recommend#

• LM Studio recommends 16 GB or more of memory on a Mac. It says 8 GB is possible with smaller models. On Windows it recommends at least 16 GB of memory and at least 4 GB of graphics memory. See its system requirements.
• Ollama recommends 8 GB of available graphics memory, or unified memory on a Mac, for its example model. See the quickstart.

These are recommendations from the app makers for their own examples. They are not test results. Our tiers are the place to see what actually happens on real jobs.

How to check what your laptop has#

• Mac: Open the Apple menu and choose About This Mac. To check whether you can add memory later, hold the Option key, open the Apple menu and choose System Information. Under Hardware, click Memory and look at "Upgradeable Memory". See Apple's guide.
• Windows: Open Settings, then System, then About. Under Device specifications you will see the processor and the installed RAM. See Microsoft's guide.
• Graphics card memory: Check the maker's specification sheet for your exact model.

What this means for your laptop#

• Buying: Memory is the part you may not be able to change later. On a Mac, check the "Upgradeable Memory" line. Our guide on buying a refurbished laptop covers this.
• Already own one: Find your tier above, open its page, and check the verdicts for your jobs.
• Not sure: Open the laptop list and filter by tier. Or ask us about our setup service, where we check the machine for you.