Mac Studio M3 Ultra 96GB vs NVIDIA GeForce RTX 5090

For running models locally, priced in Germany. At 128k context the first runs 37 models fully in memory, the second 24.

In short

The Mac Studio M3 Ultra 96GB has more memory for models: 96 GB against 32 GB. At 128k context the Mac Studio M3 Ultra 96GB runs 37 of the models we track fully in memory and the NVIDIA GeForce RTX 5090 24; only the Mac Studio M3 Ultra 96GB holds Qwen 3.5 122B-A10B. On the models both run fully in memory, the NVIDIA GeForce RTX 5090 is about 2.2× as fast as the Mac Studio M3 Ultra 96GB (estimated, median across models). In Germany the Mac Studio M3 Ultra 96GB costs 5.999 € at eBay and the NVIDIA GeForce RTX 5090 6.028 € at Amazon: the Mac Studio M3 Ultra 96GB is 29 € cheaper.

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

Specifications

SpecMac Studio M3 Ultra 96GBNVIDIA GeForce RTX 5090
ChipApple M3 Ultra (80-core GPU)–
Memory bus1024-bit512-bit
Memory bandwidth819 GB/s1,792 GB/s
FP16 compute28 TFLOPS209 TFLOPS
SoftwareMetalCUDA
Unified memory96 GB–
GPU can use by default72 GB–
GPU can use, tuned88 GB–
Operating systemmacos–
ProcessorM3 Ultra 32-core–
Storage1000 GB–
GPU–GeForce RTX 5090
Memory–32 GB GDDR7
Board power–575 W
PCIe generation–Gen 5
PCIe lanes–x16

What each runs

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

ModelMac Studio M3 Ultra 96GBNVIDIA GeForce RTX 5090
Qwen 3.8 27B14tok/sQ8_036tok/sUD-Q6_K
Qwen 3.5 122B-A10B41tok/sQ3_K_Moffload
Qwen 3.6 27B14tok/sQ8_036tok/sQ6_K
Qwen3 Coder Next49tok/sQ6_Koffload
Nemotron 3 Super 120B-A12B44tok/sUD-Q3_K_Moffload
gpt-oss 120B40tok/sMXFP4won't run
Gemma 4 31B12tok/sQ8_037tok/sQ4_K_M
Mistral Small 4 119B51tok/sUD-Q3_K_Moffload
Qwen 3.6 35B-A3B55tok/sQ8_0128tok/sUD-Q6_K
Qwen 3.5 27B14tok/sQ8_036tok/sQ6_K
Gemma 4 26B-A4B45tok/sQ8_095tok/sQ8_0
GLM 4.5 Air10tok/sQ2_Kwon't run
Qwen3 Next 80B-A3B56tok/sQ6_Koffload
Qwen 3.5 35B-A3B55tok/sQ8_0129tok/sQ6_K
GLM 4.7 Flash26tok/sQ8_060tok/sQ6_K
Devstral Small 2 24B11tok/sQ8_0offload
Nemotron 3.5 Lightning 30B-A3B76tok/sUD-Q6_K_XLoffload
Kimi Linear 48B-A3B75tok/sQ8_0227tok/sQ4_K_M
gpt-oss 20B57tok/sMXFP4118tok/sMXFP4
Qwen3 Coder 30B-A3B20tok/sQ8_045tok/sQ4_K_M
Qwen3 32B8tok/sQ8_0offload
Gemma 4 12B35tok/sQ8_073tok/sQ8_0
Qwen 3.5 9B38tok/sQ8_080tok/sQ8_0
Gemma 3 27B13tok/sQ8_037tok/sQ5_K_M
Mistral Small 3.2 24B11tok/sQ8_0offload
DeepSeek R1 Distill Qwen 32B8tok/sQ8_0offload
Ministral 3 14B15tok/sQ8_036tok/sQ5_K_M
Qwen3 14B14tok/sQ8_035tok/sQ5_K_M
DeepSeek R1 0528 Qwen3 8B19tok/sQ8_040tok/sQ8_0
Qwen3 8B19tok/sQ8_040tok/sQ8_0
Phi-4 14B13tok/sQ8_0offload
Qwen 3.5 4B60tok/sQ8_0126tok/sQ8_0
Ministral 3 8B20tok/sQ8_041tok/sQ8_0
Granite 4.2 8B17tok/sQ8_036tok/sQ8_0
Gemma 4 E4B44tok/sQ8_093tok/sQ8_0
LFM2.5 8B-A1B43tok/sQ8_089tok/sQ8_0
Llama 3.1 8B22tok/sQ6_K47tok/sQ6_K
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
Llama 3.3 70Bwon't runwon't run

Questions

Which is better for local AI, the Mac Studio M3 Ultra 96GB or the NVIDIA GeForce RTX 5090?
Choose the Mac Studio M3 Ultra 96GB to run bigger models. Choose the NVIDIA GeForce RTX 5090 for speed. Choose the Mac Studio M3 Ultra 96GB to spend less.
Which can run bigger models, the Mac Studio M3 Ultra 96GB or the NVIDIA GeForce RTX 5090?
At 128k context the Mac Studio M3 Ultra 96GB runs 37 of the models we track fully in memory and the NVIDIA GeForce RTX 5090 24; only the Mac Studio M3 Ultra 96GB holds Qwen 3.5 122B-A10B.
Which is faster for local AI, the Mac Studio M3 Ultra 96GB or the NVIDIA GeForce RTX 5090?
On the models both run fully in memory, the NVIDIA GeForce RTX 5090 is about 2.2× as fast as the Mac Studio M3 Ultra 96GB (estimated, median across models).
What is the price difference between the Mac Studio M3 Ultra 96GB and the NVIDIA GeForce RTX 5090?
In Germany the Mac Studio M3 Ultra 96GB costs 5.999 € at eBay and the NVIDIA GeForce RTX 5090 6.028 € at Amazon: the Mac Studio M3 Ultra 96GB is 29 € cheaper.

More comparisons