Home / Machines / M1 Max / M2 Max / M3 Max

What AI can a Mac with M1 Max / M2 Max / M3 Max run?

Every memory option compared, 400 GB/s memory bandwidth. Updated 2 November 2026.

Quick verdict

M1 Max / M2 Max / M3 Max are sold with 32, 36, 48, 64, 96, 128 GB of memory. Counting the 31 models tracked here: 32 GB runs 26 models well; 36 GB runs 26 models well; 48 GB runs 27 models well; 64 GB runs 27 models well; 96 GB runs 29 models well; 128 GB runs 29 models well. Memory bandwidth is 400 GB/s, about 3.3 times the memory speed of a base M4 (120 GB/s), so the same model generates words proportionally faster. Apple Silicon shares one pool of memory between the processor and the GPU, and macOS lets the GPU use about two thirds of it on Macs up to 36 GB and about three quarters above that, so the memory you buy decides which models load at all. Not every size is sold on every chip: M1 Max: 32, 64 GB; M2 Max: 32, 64, 96 GB; M3 Max: 36, 48, 64, 96, 128 GB.

Which memory should you choose?

Smallest configuration for popular models

Every model, every memory option

Model32 GB36 GB48 GB64 GB96 GB128 GB
Qwen 3.5 122B A10BNoNoNoNoTight 18/sGreat 20/s
gpt-oss 120BNoNoNoNoGreat 38/sGreat 38/s
Llama 3.3 70BNoNoNoSlow 5/sSlow 5/sSlow 5/s
Wan 2.2 14BSlow about 28 minSlow about 28 minGreat about 14 minGreat about 14 minGreat about 14 minGreat about 14 min
LTX-2.3Great about 2 minGreat about 2 minGreat about 2 minGreat about 2 minGreat about 2 minGreat about 2 min
Qwen 3.6 35B A3BTight 54/sTight 54/sGreat 60/sGreat 60/sGreat 60/sGreat 60/s
Qwen 3.5 35B A3BTight 56/sGreat 63/sGreat 63/sGreat 63/sGreat 63/sGreat 63/s
Gemma 4 31BTight 10/sGood 11/sGood 11/sGood 11/sGood 11/sGood 11/s
DeepSeek R1 32BTight 10/sGood 11/sGood 11/sGood 11/sGood 11/sGood 11/s
Qwen3 Coder 30BGreat 58/sGreat 58/sGreat 58/sGreat 58/sGreat 58/sGreat 58/s
Qwen 3.6 27BGreat 11/sGreat 11/sGreat 11/sGreat 11/sGreat 11/sGreat 11/s
Gemma 4 26B A4BGreat 45/sGreat 45/sGreat 45/sGreat 45/sGreat 45/sGreat 45/s
Z-Image Turbo 6BGreat about 27 sGreat about 27 sGreat about 27 sGreat about 27 sGreat about 27 sGreat about 27 s
Wan 2.2 5BGreat about 7 minGreat about 7 minGreat about 7 minGreat about 7 minGreat about 7 minGreat about 7 min
gpt-oss 20BGreat 48/sGreat 48/sGreat 48/sGreat 48/sGreat 48/sGreat 48/s
ACE-Step 1.5Great about 1 minGreat about 1 minGreat about 1 minGreat about 1 minGreat about 1 minGreat about 1 min
FLUX.2 klein 4BGreat about 19 sGreat about 19 sGreat about 19 sGreat about 19 sGreat about 19 sGreat about 19 s
Phi-4 14BGreat 23/sGreat 23/sGreat 23/sGreat 23/sGreat 23/sGreat 23/s
Gemma 4 12BGreat 26/sGreat 26/sGreat 26/sGreat 26/sGreat 26/sGreat 26/s
Stable Diffusion XLGreat about 11 sGreat about 11 sGreat about 11 sGreat about 11 sGreat about 11 sGreat about 11 s
Qwen3 TTS 1.7BGreat about 30 sGreat about 30 sGreat about 30 sGreat about 30 sGreat about 30 sGreat about 30 s
Gemma 4 E4BGreat 66/sGreat 66/sGreat 66/sGreat 66/sGreat 66/sGreat 66/s
Qwen 3.5 9BGreat 32/sGreat 32/sGreat 32/sGreat 32/sGreat 32/sGreat 32/s
Qwen3 VL 8BGreat 35/sGreat 35/sGreat 35/sGreat 35/sGreat 35/sGreat 35/s
DeepSeek R1 8BGreat 41/sGreat 41/sGreat 41/sGreat 41/sGreat 41/sGreat 41/s
Qwen2.5 Coder 7BGreat 45/sGreat 45/sGreat 45/sGreat 45/sGreat 45/sGreat 45/s
Gemma 4 E2BGreat 105/sGreat 105/sGreat 105/sGreat 105/sGreat 105/sGreat 105/s
Whisper large-v3 turboGreat about 42 sGreat about 42 sGreat about 42 sGreat about 42 sGreat about 42 sGreat about 42 s
Qwen 3.5 4BGreat 64/sGreat 64/sGreat 64/sGreat 64/sGreat 64/sGreat 64/s
Kokoro TTSGreat about 6 sGreat about 6 sGreat about 6 sGreat about 6 sGreat about 6 sGreat about 6 s

Text models show words a second. Image, video and voice models show the time per job (image, 4 second clip, minute of speech or similar).

Each configuration in detail

As an Amazon Associate I earn from qualifying purchases.

M1 Max / M2 Max / M3 Max with 32 GB

About 30 GB is usable by the GPU. The largest chat or coding model that fits is Qwen 3.6 35B A3B (24.6 GB), generating about 54 words a second with 5.8 GB to spare. The quickest is Gemma 4 E2B at about 105 words a second. Closest miss: Llama 3.3 70B, which needs about 45.6 GB (15 GB more than this configuration can give the GPU).

Find M2 Max Macs with 32 GB on Amazon

M1 Max / M2 Max / M3 Max with 36 GB

About 34 GB is usable by the GPU. The largest chat or coding model that fits is Qwen 3.6 35B A3B (24.6 GB), generating about 54 words a second with 9.6 GB to spare. The quickest is Gemma 4 E2B at about 105 words a second. Closest miss: Llama 3.3 70B, which needs about 45.6 GB (11 GB more than this configuration can give the GPU).

Find M3 Max Macs with 36 GB on Amazon

M1 Max / M2 Max / M3 Max with 48 GB

About 46 GB is usable by the GPU. The largest chat or coding model that fits is Qwen 3.6 35B A3B (24.6 GB), generating about 60 words a second with 21 GB to spare. The quickest is Gemma 4 E2B at about 105 words a second. Closest miss: Llama 3.3 70B, which needs about 45.6 GB (1 GB more than this configuration can give the GPU).

Find M3 Max Macs with 48 GB on Amazon

M1 Max / M2 Max / M3 Max with 64 GB

About 61 GB is usable by the GPU. The largest chat or coding model that fits is Qwen 3.6 35B A3B (24.6 GB), generating about 60 words a second with 36.2 GB to spare. The quickest is Gemma 4 E2B at about 105 words a second. Closest miss: gpt-oss 120B, which needs about 68.8 GB (8 GB more than this configuration can give the GPU).

Find M3 Max Macs with 64 GB on Amazon

M1 Max / M2 Max / M3 Max with 96 GB

About 91 GB is usable by the GPU. The largest chat or coding model that fits is Qwen 3.5 122B A10B (85.5 GB), generating about 18 words a second with 5.7 GB to spare. The quickest is Gemma 4 E2B at about 105 words a second. Closest miss: DeepSeek V3 671B, which needs about 424.7 GB (334 GB more than this configuration can give the GPU).

Find M3 Max Macs with 96 GB on Amazon

M1 Max / M2 Max / M3 Max with 128 GB

About 122 GB is usable by the GPU. The largest chat or coding model that fits is Qwen 3.5 122B A10B (85.5 GB), generating about 20 words a second with 36.1 GB to spare. The quickest is Gemma 4 E2B at about 105 words a second. Closest miss: DeepSeek V3 671B, which needs about 424.7 GB (303 GB more than this configuration can give the GPU).

Find M3 Max Macs with 128 GB on Amazon

Questions

How much memory do I need on M1 Max / M2 Max / M3 Max for local AI?

For everyday chat, 32 GB runs 26 of 31 models well. 96 GB to 128 GB unlocks nothing new in this list. The extra memory only gives longer conversations and more room for other apps.

Can M1 Max / M2 Max / M3 Max run Gemma 4 31B?

Yes, from 32 GB of memory, at about 10 words a second.

How fast is M1 Max / M2 Max / M3 Max for local AI?

32 GB runs gpt-oss 20B at about 48 words a second; 36 GB runs gpt-oss 20B at about 48 words a second; 48 GB runs gpt-oss 20B at about 48 words a second; 64 GB runs gpt-oss 20B at about 48 words a second; 96 GB runs gpt-oss 20B at about 48 words a second; 128 GB runs gpt-oss 20B at about 48 words a second.

What an upgrade would unlock

As an Amazon Associate I earn from qualifying purchases.

Mac with M5 Max, 128 GB

Unlocks 1 more: Llama 3.3 70B.

See prices
Too slow at home? Rent a cloud GPU by the hour and run any of these models.Try RunPod