← Every GPUApple · 36 or 48 or 64 or 128 GB unified memory · 546 GB/s

What can the Apple M4 Max run?

With 36 GB, 8 of 12 popular open models fit. The best pick is Qwen3 30B-A3B at about 96 tok/s, faster than you can read.

ollama run qwen3:30b

First time? Install Ollama, then paste that into a terminal. Is this your machine? Check it on the homepage.

36 GB unified memory · about 24 GB for models

Every model the Apple M4 Max 36 GB can run

8 of 12 fit. Best pick: Qwen3 30B-A3B.

  • Qwen3 4BQuick everyday chat and summaries
    ~120 tok/s
    Runs greatqwen3:4b
  • Qwen3 8BA solid all-rounder for chat and writing
    ~58 tok/s
    Runs greatqwen3:8b
  • Gemma 3 12BReads images as well as text
    ~37 tok/s
    Runs greatgemma3:12b
  • Qwen3 14BNoticeably smarter at reasoning and code
    ~32 tok/s
    Runs greatqwen3:14b
  • gpt-oss 20BOpenAI's open model. Fast, strong reasoning
    ~96 tok/s
    Runs greatgpt-oss:20b
  • Gemma 3 27BGoogle's best open model for its size
    ~18 tok/s
    Usablegemma3:27b
  • Qwen3 30B-A3B30B smarts at small-model speed
    ~96 tok/s
    Runs greatqwen3:30b
  • Qwen3 32BStrong coding and long answers
    ~15 tok/s
    Usableqwen3:32b
  • Llama 3.3 70BClose to the big cloud models
    needs 45 GB
    Too bigllama3.3:70b
  • gpt-oss 120BOpenAI's largest open model
    needs 67 GB
    Too biggpt-oss:120b
  • Qwen3 235B-A22BAlibaba's flagship open model, for multi-GPU rigs
    needs 144 GB
    Too bigqwen3:235b
  • DeepSeek R1 671BThe full DeepSeek R1, for the biggest rigs
    needs 406 GB
    Too bigdeepseek-r1:671b
48 GB unified memory · about 36 GB for models

Every model the Apple M4 Max 48 GB can run

8 of 12 fit. Best pick: Qwen3 30B-A3B.

  • Qwen3 4BQuick everyday chat and summaries
    ~120 tok/s
    Runs greatqwen3:4b
  • Qwen3 8BA solid all-rounder for chat and writing
    ~58 tok/s
    Runs greatqwen3:8b
  • Gemma 3 12BReads images as well as text
    ~37 tok/s
    Runs greatgemma3:12b
  • Qwen3 14BNoticeably smarter at reasoning and code
    ~32 tok/s
    Runs greatqwen3:14b
  • gpt-oss 20BOpenAI's open model. Fast, strong reasoning
    ~96 tok/s
    Runs greatgpt-oss:20b
  • Gemma 3 27BGoogle's best open model for its size
    ~18 tok/s
    Usablegemma3:27b
  • Qwen3 30B-A3B30B smarts at small-model speed
    ~96 tok/s
    Runs greatqwen3:30b
  • Qwen3 32BStrong coding and long answers
    ~15 tok/s
    Usableqwen3:32b
  • Llama 3.3 70BClose to the big cloud models
    needs 45 GB
    Too bigllama3.3:70b
  • gpt-oss 120BOpenAI's largest open model
    needs 67 GB
    Too biggpt-oss:120b
  • Qwen3 235B-A22BAlibaba's flagship open model, for multi-GPU rigs
    needs 144 GB
    Too bigqwen3:235b
  • DeepSeek R1 671BThe full DeepSeek R1, for the biggest rigs
    needs 406 GB
    Too bigdeepseek-r1:671b
64 GB unified memory · about 48 GB for models

Every model the Apple M4 Max 64 GB can run

9 of 12 fit. Best pick: Qwen3 30B-A3B.

  • Qwen3 4BQuick everyday chat and summaries
    ~120 tok/s
    Runs greatqwen3:4b
  • Qwen3 8BA solid all-rounder for chat and writing
    ~58 tok/s
    Runs greatqwen3:8b
  • Gemma 3 12BReads images as well as text
    ~37 tok/s
    Runs greatgemma3:12b
  • Qwen3 14BNoticeably smarter at reasoning and code
    ~32 tok/s
    Runs greatqwen3:14b
  • gpt-oss 20BOpenAI's open model. Fast, strong reasoning
    ~96 tok/s
    Runs greatgpt-oss:20b
  • Gemma 3 27BGoogle's best open model for its size
    ~18 tok/s
    Usablegemma3:27b
  • Qwen3 30B-A3B30B smarts at small-model speed
    ~96 tok/s
    Runs greatqwen3:30b
  • Qwen3 32BStrong coding and long answers
    ~15 tok/s
    Usableqwen3:32b
  • Llama 3.3 70BClose to the big cloud models
    ~7 tok/s
    Slowllama3.3:70b
  • gpt-oss 120BOpenAI's largest open model
    needs 67 GB
    Too biggpt-oss:120b
  • Qwen3 235B-A22BAlibaba's flagship open model, for multi-GPU rigs
    needs 144 GB
    Too bigqwen3:235b
  • DeepSeek R1 671BThe full DeepSeek R1, for the biggest rigs
    needs 406 GB
    Too bigdeepseek-r1:671b
128 GB unified memory · about 96 GB for models

Every model the Apple M4 Max 128 GB can run

10 of 12 fit. Best pick: gpt-oss 120B.

  • Qwen3 4BQuick everyday chat and summaries
    ~120 tok/s
    Runs greatqwen3:4b
  • Qwen3 8BA solid all-rounder for chat and writing
    ~58 tok/s
    Runs greatqwen3:8b
  • Gemma 3 12BReads images as well as text
    ~37 tok/s
    Runs greatgemma3:12b
  • Qwen3 14BNoticeably smarter at reasoning and code
    ~32 tok/s
    Runs greatqwen3:14b
  • gpt-oss 20BOpenAI's open model. Fast, strong reasoning
    ~96 tok/s
    Runs greatgpt-oss:20b
  • Gemma 3 27BGoogle's best open model for its size
    ~18 tok/s
    Usablegemma3:27b
  • Qwen3 30B-A3B30B smarts at small-model speed
    ~96 tok/s
    Runs greatqwen3:30b
  • Qwen3 32BStrong coding and long answers
    ~15 tok/s
    Usableqwen3:32b
  • Llama 3.3 70BClose to the big cloud models
    ~7 tok/s
    Slowllama3.3:70b
  • gpt-oss 120BOpenAI's largest open model
    ~66 tok/s
    Runs greatgpt-oss:120b
  • Qwen3 235B-A22BAlibaba's flagship open model, for multi-GPU rigs
    needs 144 GB
    Too bigqwen3:235b
  • DeepSeek R1 671BThe full DeepSeek R1, for the biggest rigs
    needs 406 GB
    Too bigdeepseek-r1:671b
Common questions

Can the Apple M4 Max run it?

Can the Apple M4 Max run Llama 3.3 70B?

Yes, with 64 GB or more. Llama 3.3 70B needs about 45 GB and the Apple M4 Max with 64 GB of unified memory leaves about 48 GB for models. Expect about 7 tok/s, slow, but it works. The 36 and 48 GB versions don't have room.

Can the Apple M4 Max run gpt-oss 120B?

Yes, with 128 GB or more. gpt-oss 120B needs about 67 GB and the Apple M4 Max with 128 GB of unified memory leaves about 96 GB for models. Expect about 66 tok/s, faster than you can read. The 36, 48 and 64 GB versions don't have room.

Can the Apple M4 Max run Qwen3 235B-A22B?

No. Qwen3 235B-A22B needs about 144 GB and the Apple M4 Max has at most 128 GB of unified memory, of which about 96 GB is usable for models. The most capable model it runs well is gpt-oss 120B, at about 66 tok/s.

Measured, not estimated

Real runs on the Apple M4 Max

Loading measured runs…

How we estimate

Generating a token reads the model's active weights from memory once, so speed is roughly memory bandwidth (546 GB/s) divided by the active weight size, with efficiency calibrated on our own rack. Sizes are Ollama's default 4-bit downloads plus about 1.2 GB for context. macOS lets the GPU use about two thirds to three quarters of unified memory, so that is what we count. Real results are usually within about a third. Measured runs are on the leaderboard.

Want something bigger?

The Ryzen AI Max+ 395 mini PC, 128 GB runs gpt-oss 120B at about 31 tok/s, from $3,400 used.

See the builds

Compare

Similar cards

Every GPU we track

Head to head

The Apple M4 Max against its rivals

Every comparison