Hardware

Apple M5 Max (40-core GPU)

Apple M5 Max (40-core GPU) has 48 / 64 / 128 GB Unified and 614 GB/s. See which open models fit, at which quant, and roughly how fast. Figures below use the default memory (48 GB). Updated 2026-10-17.

Memory48 / 64 / 128 GB Unified

Bandwidth614 GB/s

Kindmac

Year2026

Apple: up to 614 GB/s and 128GB for the 40-core GPU M5 Max. Source.

RSS for new models on this device

Models that fit

Model	Verdict	Quant	Context	tok/s
Qwen3 Coder Next 79.7B	Tight fit	UD-Q3_K_XL	8K	49–66
Qwen3 Next 80B A3B Thinking 79.7B	Tight fit	UD-Q3_K_XL	8K	158–215
Qwen3 Next 80B A3B Instruct 79.7B	Tight fit	UD-Q3_K_XL	8K	157–214
GLM 4 32B 0414 32.6B	Tight fit	Q8_0	8K	9.6–13
Mistral Small 4 119B 2603 119B	Tight fit	UD-IQ2_XXS	8K	52–71
Qwen3.8 27B 27.3B	Runs	UD-Q8_K_L	8K	12–16
Kwaipilot KAT Coder V2.5 Dev 34.7B	Tight fit	Q8_0	8K	49–67
Qwen3.8 35B A3B Distill 35.5B	Runs	Q6_K	8K	128–175
Qwen3.6 35B A3B 34.7B	Tight fit	Q8_0	8K	100–137
Qwen3.5 35B A3B 34.7B	Tight fit	Q8_0	8K	100–137
Llama 4 Scout 17B 16E Instruct 108B	Tight fit	UD-IQ1_M	8K	37–51
NVIDIA Nemotron 3.5 Lightning 30B A3B 32.9B	Tight fit	Q8_0	8K	93–127
Devstral Small 2 24B Instruct 2512 23.6B	Runs	UD-Q8_K_XL	8K	11–15
Mistral Small 3.2 24B Instruct 2506 23.6B	Runs	UD-Q8_K_XL	8K	11–15
Magistral Small 2509 23.6B	Runs	UD-Q8_K_XL	8K	11–15
Nemotron 3 Nano 30B A3B 31.6B	Tight fit	Q8_0	8K	93–127
NVIDIA Nemotron 3 Nano Omni 30B A3B Reasoning 31.6B	Tight fit	UD-Q8_K_XL	8K	89–121
GLM 4.7 Flash 29.9B	Tight fit	UD-Q8_K_XL	8K	42–57
Qwen3 Coder 30B A3B Instruct 30.5B	Tight fit	UD-Q8_K_XL	8K	78–106
Qwen3 30B A3B Instruct 2507 30.5B	Tight fit	UD-Q8_K_XL	8K	78–106
Qwen3 VL 30B A3B Instruct 30.5B	Tight fit	UD-Q8_K_XL	8K	78–106
Qwen3 30B A3B Thinking 2507 30.5B	Tight fit	UD-Q8_K_XL	8K	78–106
gemma 4 26B A4B it 25.2B	Runs	UD-Q8_K_XL	8K	53–72
LFM2.5 24B A2B 24.0B	Runs	Q8_0	8K	121–165
GLM 4.7 Flash REAP 23B A3B 23.0B	Runs	UD-Q8_K_XL	8K	84–114
gpt oss 20b 20.9B	Runs great	UD-Q8_K_XL	8K	84–114
Qwen2.5 14B Instruct 14.8B	Runs	F16	8K	11–15
Qwen3 14B 14.8B	Runs	BF16	8K	11–15
NousResearch Hermes 4 14B 14.8B	Runs	BF16	8K	11–15
DeepSeek R1 Distill Qwen 14B 14.8B	Runs	F16	8K	11–15
Qwen2.5 Coder 14B Instruct 128K 14.8B	Runs	F16	8K	11–15
phi 4 14.7B	Runs	F16	8K	11–15
Ministral 3 14B Instruct 2512 13.5B	Runs	BF16	8K	12–16
Ministral 3 14B Reasoning 2512 13.5B	Runs	BF16	8K	12–16
gemma 4 12b it 11.9B	Runs	BF16	8K	12–17
GLM 4.6V Flash 9.4B	Runs great	BF16	8K	18–24
GLM Z1 9B 0414 9.4B	Runs great	BF16	8K	18–24
GLM 4 9B 0414 9.4B	Runs great	BF16	8K	18–24
GLM 4.1V 9B Thinking 9.4B	Runs great	BF16	8K	18–24
Qwen3.5 9B 9.0B	Runs great	BF16	8K	19–25

90 more models in the catalog do not fit this default profile, or ranked below the first 40 that do.

Example: Tight fit fully in unified memory on your Apple M5 Max (40-core GPU) at UD-Q3_K_XL, 8K context, about 49–66 tok/s. Weights 33.8 GB plus 0.75 GB of KV cache — 35.2 GB of 36.0 GB usable. Weights shown in the verdict use GiB (the same 1024³ counting as GPU memory).