Skip to content
mlx.app

M5 Max · 36 GB

614 GB/s memory bandwidth · 25.9 GB usable for models · 10.1 GB left to macOSunverified specs

The short answer

32 models run comfortably and 0 more will run tight at 8K context. The largest that fits is Qwen2.5-VL 32B Instruct at q4, generating around 27.5 tokens per second.

macOS10.1 GBweights18.6 GBKV cache2.1 GBheadroom5.2 GB

Every model on this Mac

ModelSizeBest quantVerdictSpeed est.
Qwen2.5 0.5B Instruct490Mbf16FITS · 24.8 GB FREE520 tok/s
BGE-M3567Mbf16FITS · 24.8 GB FREE449 tok/s
Qwen3 0.6B600Mbf16FITS · 23.8 GB FREE425 tok/s
Parakeet TDT 0.6B v2600Mbf16FITS · 24.7 GB FREE425 tok/s
Llama 3.2 1B Instruct1.2Bbf16FITS · 23.2 GB FREE205 tok/s
Whisper Large v31.6Bbf16FITS · 22.8 GB FREE164 tok/s
Qwen3 1.7B1.7Bbf16FITS · 21.6 GB FREE150 tok/s
Llama 3.2 3B Instruct3.2Bbf16FITS · 19.0 GB FREE79.4 tok/s
Qwen3 4B4Bq8FITS · 20.5 GB FREE120 tok/s
Qwen3 Embedding 4B4Bq8FITS · 20.5 GB FREE120 tok/s
Mistral 7B Instruct v0.37.3Bq8FITS · 17.1 GB FREE65.7 tok/s
Qwen2.5 7B Instruct7.6Bbf16FITS · 10.3 GB FREE33.5 tok/s
InternVL3 8B8Bq8FITS · 17.4 GB FREE60.0 tok/s
Llama 3.1 8B Instruct8.0Bbf16FITS · 8.8 GB FREE31.7 tok/s
Qwen3 8B8.2Bq8FITS · 16.0 GB FREE58.5 tok/s
Qwen2.5-VL 7B Instruct8.3Bq8FITS · 16.6 GB FREE57.8 tok/s
Gemma 2 9B IT9.2Bq8FITS · 16.1 GB FREE51.9 tok/s
Llama 3.2 11B Vision Instruct10.7Bbf16FITS · 4.5 GB FREE23.8 tok/s
Qwen2.5 14B Instruct14.7Bq8FITS · 8.7 GB FREE32.6 tok/s
Qwen3 14B14.8Bq8FITS · 8.9 GB FREE32.4 tok/s
GPT-OSS 20B21Bq4FITS · 13.7 GB FREE252 tok/s
Codestral 22B v0.122.2Bq4FITS · 11.6 GB FREE40.8 tok/s
Mistral Small 24B Instruct 250124Bq4FITS · 11.1 GB FREE37.7 tok/s
Gemma 3 27B IT27Bq4FITS · 10.7 GB FREE33.6 tok/s
Gemma 2 27B IT27.2Bq4FITS · 10.6 GB FREE33.3 tok/s
Qwen3 30B-A3B Instruct 250730.5Bq4FITS · 8.0 GB FREE275 tok/s
Qwen3 Coder 30B-A3B Instruct30.5Bq4FITS · 8.0 GB FREE275 tok/s
Qwen2.5 32B Instruct32.5Bq4FITS · 5.5 GB FREE27.9 tok/s
Qwen2.5 Coder 32B Instruct32.5Bq4FITS · 5.5 GB FREE27.9 tok/s
DeepSeek R1 Distill Qwen 32B32.5Bq4FITS · 5.5 GB FREE27.9 tok/s
Qwen3 32B32.8Bq4FITS · 5.3 GB FREE27.6 tok/s
Qwen2.5-VL 32B Instruct33Bq4FITS · 5.2 GB FREE27.5 tok/s
Mixtral 8x7B Instruct v0.146.7BOVER
Llama 3.3 70B Instruct70.6BOVER
Qwen2.5 72B Instruct72.7BOVER
Llama 3.2 90B Vision Instruct88BOVER
GPT-OSS 120B116.8BOVER
Qwen3 235B-A22B235BOVER
DeepSeek V3671BOVER