Hardware

Apple M3 Max (40-core GPU)

Apple M3 Max (40-core GPU) has 48 / 64 / 96 / 128 GB Unified and 400 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 / 96 / 128 GB Unified

Bandwidth400 GB/s

Kindmac

Year2023

40-core GPU version, 400 GB/s. Up to 128GB. 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	32–43
Qwen3 Next 80B A3B Instruct 79.7B	Tight fit	UD-Q3_K_XL	8K	102–140
Qwen3 Next 80B A3B Thinking 79.7B	Tight fit	UD-Q3_K_XL	8K	103–140
Mistral Small 4 119B 2603 119B	Tight fit	UD-IQ2_XXS	8K	34–46
Kwaipilot KAT Coder V2.5 Dev 34.7B	Tight fit	Q8_0	8K	32–44
Qwen3.8 35B A3B Distill 35.5B	Runs	Q6_K	8K	83–114
Qwen3.6 35B A3B 34.7B	Tight fit	Q8_0	8K	65–89
Qwen3.5 35B A3B 34.7B	Tight fit	Q8_0	8K	65–89
Llama 4 Scout 17B 16E Instruct 108B	Tight fit	UD-IQ1_M	8K	24–33
NVIDIA Nemotron 3.5 Lightning 30B A3B 32.9B	Tight fit	Q8_0	8K	61–83
Nemotron 3 Nano 30B A3B 31.6B	Tight fit	Q8_0	8K	61–83
NVIDIA Nemotron 3 Nano Omni 30B A3B Reasoning 31.6B	Tight fit	UD-Q8_K_XL	8K	58–79
GLM 4.7 Flash 29.9B	Tight fit	UD-Q8_K_XL	8K	27–37
Qwen3 Coder 30B A3B Instruct 30.5B	Tight fit	UD-Q8_K_XL	8K	51–69
Qwen3 30B A3B Instruct 2507 30.5B	Tight fit	UD-Q8_K_XL	8K	51–69
Qwen3 VL 30B A3B Instruct 30.5B	Tight fit	UD-Q8_K_XL	8K	51–69
Qwen3 30B A3B Thinking 2507 30.5B	Tight fit	UD-Q8_K_XL	8K	51–69
gemma 4 26B A4B it 25.2B	Runs	UD-Q8_K_XL	8K	34–47
LFM2.5 24B A2B 24.0B	Runs	Q8_0	8K	79–107
GLM 4.7 Flash REAP 23B A3B 23.0B	Runs	UD-Q8_K_XL	8K	55–74
gpt oss 20b 20.9B	Runs great	UD-Q8_K_XL	8K	55–74
GLM 4.6V Flash 9.4B	Runs great	BF16	8K	11–16
GLM Z1 9B 0414 9.4B	Runs great	BF16	8K	11–16
GLM 4 9B 0414 9.4B	Runs great	BF16	8K	11–16
GLM 4.1V 9B Thinking 9.4B	Runs great	BF16	8K	11–16
Qwen3.5 9B 9.0B	Runs great	BF16	8K	12–16
MiMo V2.6 Distill Qwen 9B 9.0B	Runs great	BF16	8K	12–16
Ministral 3 8B Instruct 2512 8.5B	Runs great	BF16	8K	12–17
Ministral 3 8B Reasoning 2512 8.5B	Runs great	BF16	8K	12–17
Qwen3 VL 8B Instruct 8.2B	Runs great	BF16	8K	13–17
DeepSeek R1 0528 Qwen3 8B 8.2B	Runs great	BF16	8K	13–17
Qwen3 8B 8.2B	Runs great	BF16	8K	13–17
Qwen3 VL 8B Thinking 8.2B	Runs great	BF16	8K	13–17
DeepSeek R1 Distill Llama 8B 8.0B	Runs great	BF16	8K	13–18
Llama 3.1 8B Instruct 8.0B	Runs great	BF16	8K	13–18
Apertus 8B Instruct 2509 8.1B	Runs great	BF16	8K	13–17
Qwen2.5 7B Instruct 7.6B	Runs great	F16	8K	14–19
Qwen2.5 VL 7B Instruct 7.6B	Runs great	BF16	8K	14–19
DeepSeek R1 Distill Qwen 7B 7.6B	Runs great	F16	8K	14–19
Qwen2.5 Coder 7B Instruct 128K 7.6B	Runs great	F16	8K	14–19

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 M3 Max (40-core GPU) at UD-Q3_K_XL, 8K context, about 32–43 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).