readygpu

Mac mini M4 Pro 48GB vs NVIDIA GeForce RTX 3090

For running models locally, priced in Germany. At 128k context the first runs 54 models fully in memory, the second 28.

In short

The Mac mini M4 Pro 48GB has more memory for models: 48 GB against 24 GB. At 128k context the Mac mini M4 Pro 48GB runs 54 of the models we track fully in memory and the NVIDIA GeForce RTX 3090 28; only the Mac mini M4 Pro 48GB holds Qwen 3.8 27B. On the models both run fully in memory, the NVIDIA GeForce RTX 3090 is about 3.3× as fast as the Mac mini M4 Pro 48GB (estimated, median across models). In Germany the Mac mini M4 Pro 48GB costs 2.699 € at eBay (used) and the NVIDIA GeForce RTX 3090 1.850 € at eBay: the NVIDIA GeForce RTX 3090 is 849 € cheaper.

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

Mac mini M4 Pro 48GB

2.699 €

eBay, used (asking)

Add to build
NVIDIA GeForce RTX 3090

1.850 €

eBay

Add to build

Specifications

SpecMac mini M4 Pro 48GBNVIDIA GeForce RTX 3090
ChipApple M4 Pro (20-core GPU)–
Memory bus256-bit384-bit
Memory bandwidth273 GB/s936 GB/s
FP16 compute8.08 TFLOPS71 TFLOPS
SoftwareMetalCUDA
Unified memory48 GB–
GPU can use by default36 GB–
GPU can use, tuned40 GB–
Operating systemmacos–
ProcessorM4 Pro 14-core–
Storage512 GB–
GPU–GeForce RTX 3090
Memory–24 GB GDDR6X
Board power–350 W
PCIe generation–Gen 4
PCIe lanes–x16

What each runs

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

ModelMac mini M4 Pro 48GBNVIDIA GeForce RTX 3090
Qwen 3.8 27B5tok/sQ6_Koffload
Qwen 3.6 27B6tok/sQ6_Koffload
Gemma 4 31B5tok/sQ6_Koffload
Qwen 3.6 35B-A3B20tok/sQ6_Koffload
Qwen 3.5 27B6tok/sQ6_Koffload
Gemma 4 26B-A4B15tok/sQ8_059tok/sQ5_K_M
Qwen 3.5 35B-A3B21tok/sQ6_Koffload
GLM 4.7 Flash10tok/sQ6_Koffload
Devstral Small 2 24B5tok/sQ4_K_Moffload
Nemotron 3.5 Lightning 30B-A3B26tok/sUD-Q6_K_XLoffload
Kimi Linear 48B-A3B33tok/sQ5_K_Moffload
gpt-oss 20B19tok/sMXFP462tok/sMXFP4
Qwen3 Coder 30B-A3B7tok/sQ5_K_Moffload
Gemma 4 12B12tok/sQ8_038tok/sQ8_0
Qwen 3.5 9B13tok/sQ8_042tok/sQ8_0
Gemma 3 27B5tok/sQ6_Koffload
Mistral Small 3.2 24B5tok/sQ4_K_Moffload
Ministral 3 14B5tok/sQ8_0offload
Qwen3 14B5tok/sQ8_0offload
DeepSeek R1 0528 Qwen3 8B6tok/sQ8_0offload
Qwen3 8B6tok/sQ8_0offload
Phi-4 14B5tok/sQ4_K_Moffload
Qwen 3.5 4B20tok/sQ8_066tok/sQ8_0
Ministral 3 8B7tok/sQ8_025tok/sQ4_K_M
Granite 4.2 8B6tok/sQ8_0offload
Gemma 4 E4B15tok/sQ8_048tok/sQ8_0
LFM2.5 8B-A1B14tok/sQ8_046tok/sQ8_0
Llama 3.1 8B8tok/sQ6_K25tok/sQ5_K_M
Qwen2.5 0.5B83tok/sQ8_0271tok/sQ8_0
Qwen3.5 0.8B73tok/sQ8_0238tok/sQ8_0
Qwen3.5 2B58tok/sQ5_K_M189tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B31tok/sQ8_0103tok/sQ8_0
Mellum2.1 12B A2.5B Thinking23tok/sQ8_076tok/sQ8_0
Spark X2.5 4B22tok/smlx-4bit73tok/smlx-4bit
MiniCPM5 2B21tok/sQ8_070tok/sQ8_0
LFM2.5 2.6B16tok/sQ8_053tok/sQ8_0
LFM2.5 VL 3B16tok/sQ8_053tok/sQ8_0
Kimi VL A3B Thinking 250615tok/sQ8_049tok/sQ8_0
Kimi VL A3B15tok/smlx-8bit48tok/smlx-8bit
GLM 4 9B 041413tok/smlx-5bit43tok/smlx-5bit
granite vision 4.1 4b12tok/sQ8_040tok/sQ8_0
granite 4.2 3b12tok/sQ8_040tok/sQ8_0
GLM 4.6V Flash12tok/sQ8_038tok/sQ8_0
DeepSeek R1 Distill Qwen 7B11tok/sQ8_037tok/sQ8_0
GLM 4.1V 9B Thinking10tok/smlx-8bit34tok/smlx-8bit
LFM2.5 24B A2B8tok/sQ8_0offload
DeepSeek R1 Distill Llama 8B7tok/sQ8_025tok/sQ5_K_M
Ministral 3 8B Reasoning 25127tok/sQ8_025tok/sQ4_K_M
Qwen3.8 27B OBLITERATED6tok/smlx-4bitoffload
Devstral Small 25075tok/sQ4_K_Moffload
Magistral Small 25065tok/sQ4_K_Moffload
Magistral Small 25075tok/sQ4_K_Moffload
Magistral Small 25095tok/sQ4_K_Moffload
DeepSeek R1 Distill Qwen 14B5tok/sQ5_K_Moffload
Qwen 3.5 122B-A10Bwon't runoffload
Mistral Small 4 119Bwon't runoffload
DeepSeek V4 Flashwon't runwon't run
GLM 5.3 Flashwon't runwon't run
Qwen 3.8 Flash Nextwon't runwon't run
MiniMax M2.7won't runwon't run
Nemotron 3 Super 120B-A12Bwon't runwon't run
Qwen3 Coder Nextwon't runwon't run
gpt-oss 120Bwon't runwon't run
GLM 4.5 Airwon't runwon't run
Qwen3 Next 80B-A3Bwon't runwon't run
Qwen3 32Bwon't runwon't run
DeepSeek R1 Distill Qwen 32Bwon't runwon't run
Llama 3.3 70Bwon't runwon't run
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek R1 Distill Llama 70Bwon't runwon't run
DeepSeek R1 Zerowon'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
Devstral 2 123B Instruct 2512won't runwon't run
GLM 4.5Vwon't runwon't run
GLM 4.6won't runwon't run
GLM 4.6Vwon't runwon't run
GLM 4.7won'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 Dev 72Bwon'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
Mistral Medium 3.5 128Bwon't runwon't run
granite 4.2 30bwon't runwon't run

Questions

Which is better for local AI, the Mac mini M4 Pro 48GB or the NVIDIA GeForce RTX 3090?
Choose the Mac mini M4 Pro 48GB to run bigger models. Choose the NVIDIA GeForce RTX 3090 for speed. Choose the NVIDIA GeForce RTX 3090 to spend less.
Which can run bigger models, the Mac mini M4 Pro 48GB or the NVIDIA GeForce RTX 3090?
At 128k context the Mac mini M4 Pro 48GB runs 54 of the models we track fully in memory and the NVIDIA GeForce RTX 3090 28; only the Mac mini M4 Pro 48GB holds Qwen 3.8 27B.
Which is faster for local AI, the Mac mini M4 Pro 48GB or the NVIDIA GeForce RTX 3090?
On the models both run fully in memory, the NVIDIA GeForce RTX 3090 is about 3.3× as fast as the Mac mini M4 Pro 48GB (estimated, median across models).
What is the price difference between the Mac mini M4 Pro 48GB and the NVIDIA GeForce RTX 3090?
In Germany the Mac mini M4 Pro 48GB costs 2.699 € at eBay (used) and the NVIDIA GeForce RTX 3090 1.850 € at eBay: the NVIDIA GeForce RTX 3090 is 849 € cheaper.

More comparisons