readygpu

Mac Studio M4 Max 64GB vs NVIDIA DGX Spark 128GB

For running models locally, priced in Canada. At 128k context the first runs 60 models fully in memory, the second 79.

In short

The NVIDIA DGX Spark 128GB has more memory for models: 128 GB against 64 GB. At 128k context the NVIDIA DGX Spark 128GB runs 79 of the models we track fully in memory and the Mac Studio M4 Max 64GB 60; only the NVIDIA DGX Spark 128GB holds DeepSeek V4 Flash. On the models both run fully in memory, the Mac Studio M4 Max 64GB is about 2.1× as fast as the NVIDIA DGX Spark 128GB (estimated, median across models).

Updated . Speeds are estimates; how they are worked out.

Mac Studio M4 Max 64GB

Not priced here

Add to build
NVIDIA DGX Spark 128GB

$11,014

Amazon

Add to build

Specifications

SpecMac Studio M4 Max 64GBNVIDIA DGX Spark 128GB
ChipApple M4 Max (40-core GPU)GB10 Grace Blackwell
Memory bus512-bit256-bit
Memory bandwidth546 GB/s273 GB/s
FP16 compute16.16 TFLOPS62 TFLOPS
SoftwareMetalCUDA
Unified memory64 GB128 GB
GPU can use by default48 GB119 GB
GPU can use, tuned56 GB122 GB
Operating systemmacosdgx-os
ProcessorM4 Max 16-core20-core Arm (GB10)
Storage1000 GB4000 GB

What each runs

Estimated decode speed with the most precise quant that fits each one est.

ModelMac Studio M4 Max 64GBNVIDIA DGX Spark 128GB
DeepSeek V4 Flashwon't run6tok/sUD-Q2_K_XL
GLM 5.3 Flashwon't run10tok/sUD-Q2_K_XL
Qwen 3.8 Flash Nextwon't run2tok/sQ4_K_M
Qwen 3.8 27B9tok/sQ8_05tok/sQ8_0
MiniMax M2.7won't run3tok/sUD-Q2_K_XL
Qwen 3.5 122B-A10B32tok/sUD-Q2_K_XL9tok/sQ6_K
Qwen 3.6 27B10tok/sQ8_05tok/sQ8_0
Gemma 4 31B8tok/sQ8_04tok/sQ8_0
Nemotron 3 Super 120B-A12Bwon't run8tok/sUD-Q6_K
Qwen3 Coder Nextwon't run14tok/sQ8_0
gpt-oss 120Bwon't run13tok/sMXFP4
Mistral Small 4 119B40tok/sUD-Q2_K_XL11tok/sUD-Q6_K
Qwen 3.6 35B-A3B37tok/sQ8_017tok/sQ8_0
Qwen 3.5 27B10tok/sQ8_05tok/sQ8_0
Gemma 4 26B-A4B30tok/sQ8_014tok/sQ8_0
GLM 4.5 Airwon't run3tok/sQ6_K
Qwen 3.5 35B-A3B37tok/sQ8_017tok/sQ8_0
GLM 4.7 Flash17tok/sQ8_08tok/sQ8_0
Devstral Small 2 24B8tok/sQ8_04tok/sQ8_0
Qwen3 Next 80B-A3Bwon't run16tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3B51tok/sUD-Q6_K_XL24tok/sUD-Q6_K_XL
Kimi Linear 48B-A3B60tok/sQ6_K24tok/sQ8_0
gpt-oss 20B38tok/sMXFP418tok/sMXFP4
Qwen3 Coder 30B-A3B13tok/sQ8_06tok/sQ8_0
Qwen3 32Bwon't run2tok/sQ8_0
Gemma 4 12B23tok/sQ8_011tok/sQ8_0
Qwen 3.5 9B26tok/sQ8_012tok/sQ8_0
Gemma 3 27B9tok/sQ8_04tok/sQ8_0
Mistral Small 3.2 24B8tok/sQ8_04tok/sQ8_0
DeepSeek R1 Distill Qwen 32Bwon't run2tok/sQ8_0
Llama 3.3 70Bwon't run1tok/sQ8_0
Ministral 3 14B10tok/sQ8_05tok/sQ8_0
Qwen3 14B10tok/sQ8_05tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B13tok/sQ8_06tok/sQ8_0
Qwen3 8B13tok/sQ8_06tok/sQ8_0
Phi-4 14B8tok/sQ8_04tok/sQ8_0
Qwen 3.5 4B40tok/sQ8_019tok/sQ8_0
Ministral 3 8B13tok/sQ8_06tok/sQ8_0
Granite 4.2 8B12tok/sQ8_06tok/sQ8_0
Gemma 4 E4B30tok/sQ8_014tok/sQ8_0
LFM2.5 8B-A1B28tok/sQ8_014tok/sQ8_0
Llama 3.1 8B15tok/sQ6_K7tok/sQ6_K
Qwen2.5 0.5B166tok/sQ8_079tok/sQ8_0
Qwen3.5 0.8B145tok/sQ8_069tok/sQ8_0
Qwen3.5 2B116tok/sQ5_K_M55tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B63tok/sQ8_030tok/sQ8_0
Mellum2.1 12B A2.5B Thinking47tok/sQ8_022tok/sQ8_0
Spark X2.5 4B45tok/smlx-4bit21tok/smlx-4bit
MiniCPM5 2B43tok/sQ8_020tok/sQ8_0
FrogNano 4B 260940tok/sQ8_019tok/sQ8_0
LFM2.5 2.6B33tok/sQ8_016tok/sQ8_0
LFM2.5 VL 3B33tok/sQ8_016tok/sQ8_0
Kimi VL A3B Thinking 250630tok/sQ8_014tok/sQ8_0
Kimi VL A3B29tok/smlx-8bit14tok/smlx-8bit
GLM 4 9B 041426tok/smlx-5bit12tok/smlx-5bit
granite vision 4.1 4b25tok/sQ8_012tok/sQ8_0
granite 4.2 3b24tok/sQ8_012tok/sQ8_0
GLM 4.6V Flash23tok/sQ8_011tok/sQ8_0
DeepSeek R1 Distill Qwen 7B23tok/sQ8_011tok/sQ8_0
GLM 4.1V 9B Thinking21tok/smlx-8bit10tok/smlx-8bit
Ministral 3 3B Instruct 251220tok/sQ8_010tok/sQ8_0
LFM2.5 24B A2B17tok/sQ8_08tok/sQ8_0
DeepSeek R1 Distill Llama 8B14tok/sQ8_07tok/sQ8_0
Ministral 3 8B Reasoning 251213tok/sQ8_06tok/sQ8_0
Qwen3.8 27B OBLITERATED12tok/smlx-4bit6tok/smlx-4bit
Muse Glimmer 30B10tok/sQ8_05tok/sQ8_0
Ministral 3 14B Reasoning 251210tok/sQ8_05tok/sQ8_0
DeepSeek R1 Distill Qwen 14B9tok/sQ8_04tok/sQ8_0
Devstral Small 25078tok/sQ8_04tok/sQ8_0
Magistral Small 25068tok/sQ8_04tok/sQ8_0
Magistral Small 25078tok/sQ8_04tok/sQ8_0
Magistral Small 25098tok/sQ8_04tok/sQ8_0
DeepSeek R1 Distill Llama 70Bwon't run1tok/sQ8_0
Devstral 2 123B Instruct 2512won't run1tok/sQ4_K_M
GLM 4.5Vwon't run3tok/sQ4_K_M
GLM 4.6Vwon't run3tok/sQ4_K_M
Kimi Dev 72Bwon't run1tok/sQ8_0
Mistral Medium 3.5 128Bwon't run1tok/sQ4_K_M
granite 4.2 30bwon't run3tok/sQ8_0
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek R1 Zerowon't runwon't run
DeepSeek V3won't runwon't run
DeepSeek V3 0324won't runwon't run
DeepSeek V3.1won't runwon't run
DeepSeek V3.1 Terminuswon't runwon't run
DeepSeek V3.2won't runwon't run
DeepSeek V4 Flash 0731won't runwon't run
DeepSeek V4 Flash Vision Expwon't runwon't run
GLM 4.5won't runwon't run
GLM 4.6won't runwon't run
GLM 4.7won't runwon't run
GLM 5won't runwon't run
GLM 5.1won't runwon't run
GLM 5.2won't runwon't run
GLM 5.3won't runwon't run
Kimi K2won't runwon't run
Kimi K2 Instruct 0905won't runwon't run
Kimi K2 Thinkingwon't runwon't run
Kimi K2.5won't runwon't run
MiniMax M2won't runwon't run
MiniMax M2.1won't runwon't run
MiniMax M2.5won't runwon't run
MiniMax M3won't runwon't run

Questions

Which is better for local AI, the Mac Studio M4 Max 64GB or the NVIDIA DGX Spark 128GB?
Choose the NVIDIA DGX Spark 128GB to run bigger models. Choose the Mac Studio M4 Max 64GB for speed.
Which can run bigger models, the Mac Studio M4 Max 64GB or the NVIDIA DGX Spark 128GB?
At 128k context the NVIDIA DGX Spark 128GB runs 79 of the models we track fully in memory and the Mac Studio M4 Max 64GB 60; only the NVIDIA DGX Spark 128GB holds DeepSeek V4 Flash.
Which is faster for local AI, the Mac Studio M4 Max 64GB or the NVIDIA DGX Spark 128GB?
On the models both run fully in memory, the Mac Studio M4 Max 64GB is about 2.1× as fast as the NVIDIA DGX Spark 128GB (estimated, median across models).

More comparisons

More hardware for local AI