Fetchy

The best Mac for running local AI models

Running a large language model on your own Mac keeps your prompts at home and costs nothing per question. Tools such as Ollama, LM Studio and llama.cpp make it a few clicks, but the Mac you pick decides which models you can run and how fast they answer.

Two numbers matter more than anything else. Unified memory decides which models fit: the whole model has to stay in memory while it runs, and on Apple silicon the GPU shares that memory with the rest of the Mac. Memory bandwidth decides speed: for every word the model writes, the Mac reads the whole model once, so a Mac that moves more gigabytes per second answers faster.

This ranking puts the Mac that runs the largest model at a usable speed first and breaks ties by how quickly it answers. Each card lists what every memory option can run, and the speeds are estimates from Apple's published memory bandwidth, not measurements.

Prices are Apple's price for each configuration, not used prices.

Ranked for this use

  1. Runs up to Llama 3.3 70B at about 17 tokens/s (estimate)

    • 96 GB unified memory
    • 1,200 GB/s memory bandwidth
    • About 150 tokens/s on Llama 3.1 8B (estimate)
    • 30-core CPU · 96 GB · 1 TB → Runs up to Llama 3.3 70B at about 17 tokens/s (estimate)Apple price: €6,599
  2. Runs up to Llama 3.3 70B at about 12 tokens/s (estimate)

    • 96 GB unified memory
    • 819 GB/s memory bandwidth
    • About 102 tokens/s on Llama 3.1 8B (estimate)
    • 28-core CPU · 96 GB · 1 TB → Runs up to Llama 3.3 70B at about 12 tokens/s (estimate)Apple price: €6,299
  3. Runs up to Llama 3.3 70B at about 11 tokens/s (estimate)

    • 64 GB unified memory
    • 800 GB/s memory bandwidth
    • About 100 tokens/s on Llama 3.1 8B (estimate)
    • 24-core CPU · 64 GB · 1 TB → Runs up to Llama 3.3 70B at about 11 tokens/s (estimate)Apple price: €4,799
  4. Runs up to Llama 3.3 70B at about 11 tokens/s (estimate)

    • 64 GB unified memory
    • 800 GB/s memory bandwidth
    • About 100 tokens/s on Llama 3.1 8B (estimate)
    • 24-core CPU · 64 GB · 1 TB → Runs up to Llama 3.3 70B at about 11 tokens/s (estimate)Apple price: €8,302
  5. Runs up to Llama 3.3 70B at about 11 tokens/s (estimate)

    • 64 GB unified memory
    • 800 GB/s memory bandwidth
    • About 100 tokens/s on Llama 3.1 8B (estimate)
    • 20-core CPU · 64 GB · 1 TB → Runs up to Llama 3.3 70B at about 11 tokens/s (estimate)Apple price: €4,599
  6. Runs up to Qwen 2.5 32B at about 19 tokens/s (estimate)

    • 64 GB unified memory
    • 614 GB/s memory bandwidth
    • About 77 tokens/s on Llama 3.1 8B (estimate)
    • 18-core CPU · 36 GB · 512 GB → Runs up to Qwen 2.5 32B at about 14 tokens/s (estimate)Apple price: €2,999
    • 18-core CPU · 64 GB · 1 TB → Runs up to Qwen 2.5 32B at about 19 tokens/s (estimate)Apple price: not available for this configuration
  7. Runs up to Qwen 2.5 32B at about 13 tokens/s (estimate)

    • 36 GB unified memory
    • 410 GB/s memory bandwidth
    • About 51 tokens/s on Llama 3.1 8B (estimate)
    • 14-core CPU · 36 GB · 512 GB → Runs up to Qwen 2.5 32B at about 13 tokens/s (estimate)Apple price: €2,999
  8. Runs up to Qwen 2.5 32B at about 13 tokens/s (estimate)

    • 32 GB unified memory
    • 400 GB/s memory bandwidth
    • About 50 tokens/s on Llama 3.1 8B (estimate)
    • 12-core CPU · 32 GB · 512 GB → Runs up to Qwen 2.5 32B at about 13 tokens/s (estimate)Apple price: €2,399
  9. Runs up to Qwen 2.5 32B at about 13 tokens/s (estimate)

    • 32 GB unified memory
    • 400 GB/s memory bandwidth
    • About 50 tokens/s on Llama 3.1 8B (estimate)
    • 10-core CPU · 32 GB · 512 GB → Runs up to Qwen 2.5 32B at about 13 tokens/s (estimate)Apple price: €2,299
  10. Mac mini M5 Pro

    Good enough

    Runs up to Qwen 2.5 14B at about 22 tokens/s (estimate)

    • 24 GB unified memory
    • 307 GB/s memory bandwidth
    • About 38 tokens/s on Llama 3.1 8B (estimate)
    • 15-core CPU · 24 GB · 512 GB → Runs up to Qwen 2.5 14B at about 22 tokens/s (estimate)Apple price: €1,999
    • 18-core CPU · 64 GB · 1 TB → Runs up to Qwen 2.5 14B at about 22 tokens/s (estimate)Apple price: not available for this configuration
  11. Mac mini M4 Pro

    Good enough

    Runs up to Qwen 2.5 14B at about 19 tokens/s (estimate)

    • 24 GB unified memory
    • 273 GB/s memory bandwidth
    • About 34 tokens/s on Llama 3.1 8B (estimate)
    • 12-core CPU · 24 GB · 512 GB → Runs up to Qwen 2.5 14B at about 19 tokens/s (estimate)Apple price: €1,899
  12. Mac mini M2 Pro

    Good enough

    Runs up to Qwen 2.5 14B at about 14 tokens/s (estimate)

    • 16 GB unified memory
    • 200 GB/s memory bandwidth
    • About 25 tokens/s on Llama 3.1 8B (estimate)
    • 10-core CPU · 16 GB · 512 GB → Runs up to Qwen 2.5 14B at about 14 tokens/s (estimate)Apple price: €1,549
  13. Mac mini M6

    Good enough

    Runs up to Qwen 2.5 14B at about 12 tokens/s (estimate)

    • 24 GB unified memory
    • 170 GB/s memory bandwidth
    • About 21 tokens/s on Llama 3.1 8B (estimate)
    • 12-core CPU · 16 GB · 256 GB → Runs up to Qwen 2.5 14B at about 11 tokens/s (estimate)Apple price: €1,049
    • 12-core CPU · 16 GB · 512 GB → Runs up to Qwen 2.5 14B at about 11 tokens/s (estimate)Apple price: not available for this configuration
    • 12-core CPU · 24 GB · 512 GB → Runs up to Qwen 2.5 14B at about 12 tokens/s (estimate)Apple price: not available for this configuration
    • 12-core CPU · 32 GB · 1 TB → Runs up to Qwen 2.5 14B at about 12 tokens/s (estimate)Apple price: not available for this configuration
  14. Mac mini M4

    Not ideal

    Runs up to Llama 3.1 8B at about 15 tokens/s (estimate)

    • 16 GB unified memory
    • 120 GB/s memory bandwidth
    • About 15 tokens/s on Llama 3.1 8B (estimate)
    • 10-core CPU · 16 GB · 256 GB → Runs up to Llama 3.1 8B at about 15 tokens/s (estimate)Apple price: €949
    • 10-core CPU · 16 GB · 512 GB → Runs up to Llama 3.1 8B at about 15 tokens/s (estimate)Apple price: not available for this configuration
    • 10-core CPU · 24 GB · 512 GB → Runs up to Llama 3.1 8B at about 15 tokens/s (estimate)Apple price: not available for this configuration
  15. Mac mini M2

    Not ideal

    Llama 3.1 8B does not fit in 8 GB of memory, so this configuration cannot run it

    • 8 GB unified memory
    • 100 GB/s memory bandwidth
    • 8-core CPU · 8 GB · 256 GB → Llama 3.1 8B does not fit in 8 GB of memory, so this configuration cannot run itApple price: €699
    • 8-core CPU · 8 GB · 512 GB → Llama 3.1 8B does not fit in 8 GB of memory, so this configuration cannot run itApple price: €929
    • 8-core CPU · 8 GB · 256 GB → unknownApple price: €799
    • 8-core CPU · 8 GB · 512 GB → unknownApple price: €1,029

How we judge

Which models fit. We assume 4-bit quantised models, the usual choice for running at home, which take about 0.6 GB of memory per billion parameters. We keep 6 GB free for macOS and your other apps. A model fits if the rest of the memory is enough for it.

How fast it answers (estimate). Writing each token means reading the whole model from memory, so we divide Apple's published memory bandwidth by the size of the model and assume 60% of that peak is reached in practice. Real speed depends on the app, the model and the length of the conversation. Where a published measurement exists for a chip and model, it replaces the estimate.

Usable means at least 10 tokens per second, roughly as fast as you read. We check these reference models: Llama 3.1 8B, Qwen 2.5 14B, Qwen 2.5 32B, Llama 3.3 70B, Mistral Large 123B, and Llama 3.1 405B.

Great: runs a 70B model or larger at a usable speed. Good enough: runs at least a 14B model at a usable speed. Not ideal: only smaller models. Unknown: Apple publishes no memory bandwidth for that Mac, so we don't guess.

Memory and bandwidth come from Apple's technical specifications for each Mac. Only the configurations listed here are ranked; most Macs could be ordered with more memory.

Questions and answers

How much memory do I need to run Llama 70B on a Mac?

By our sizing, a 4-bit 70B model takes about 42 GB, and with room for macOS you need around 48 GB in total. In practice that means a Mac with 64 GB of unified memory or more. Memory alone is not enough, though: to answer at a usable speed, it also needs the memory bandwidth of an Ultra chip.

Is 16 GB enough to run AI models on a Mac?

For smaller models, yes. An 8B model such as Llama 3.1 8B needs about 10.8 GB with macOS included, so it fits in 16 GB and is useful for chat, summaries and simple coding help. A 14B model still just fits. A 32B model needs about 25.2 GB, so 32 GB, and the largest open models need 64 GB or more.

Why does memory bandwidth matter more than the chip for local AI?

Because generating text is mostly waiting for memory. For every token, the Mac reads the whole model once, so tokens per second is roughly bandwidth divided by model size. A Max or Ultra chip with several hundred GB/s answers far faster than a base chip with the same amount of memory. Extra GPU cores help most while the model reads your prompt.

Mac mini or Mac Studio for local AI?

A Mac mini with a Pro chip and enough memory runs 8B and 14B models comfortably. For 32B models and up, the Max and Ultra chips in the Mac Studio have the memory bandwidth to keep them usable, and in our ranking only Ultra chips run 70B models at a usable speed.

Which apps run AI models on a Mac?

Ollama and LM Studio are the easiest start: download a model and chat. Under the hood they use engines such as llama.cpp and Apple's MLX, which run the model on the Mac's GPU. A model's file size is a good guide to how much memory it needs.

Are the speeds on this page measured?

Not yet. They are estimates from Apple's published memory bandwidth, assuming the Mac reaches 60% of its peak, and they are labelled as estimates. Long conversations and large prompts run slower than the estimate. When a published measurement exists for a chip and model, we show it instead.