Mac Studio M3 Ultra 96GB vs NVIDIA DGX Spark 128GB
For running models locally, priced in Germany. At 128k context the first runs 37 models fully in memory, the second 42.
In short
The NVIDIA DGX Spark 128GB has more memory for models: 128 GB against 96 GB. At 128k context the NVIDIA DGX Spark 128GB runs 42 of the models we track fully in memory and the Mac Studio M3 Ultra 96GB 37; only the NVIDIA DGX Spark 128GB holds DeepSeek V4 Flash. On the models both run fully in memory, the Mac Studio M3 Ultra 96GB is about 3.2× as fast as the NVIDIA DGX Spark 128GB (estimated, median across models). In Germany the Mac Studio M3 Ultra 96GB costs 5.999 € at eBay and the NVIDIA DGX Spark 128GB 8.313 € at eBay: the Mac Studio M3 Ultra 96GB is 2.314 € cheaper.
Updated . Speeds are estimates; how they are worked out.
Specifications
| Spec | Mac Studio M3 Ultra 96GB | NVIDIA DGX Spark 128GB |
|---|---|---|
| Chip | Apple M3 Ultra (80-core GPU) | GB10 Grace Blackwell |
| Memory bus | 1024-bit | 256-bit |
| Memory bandwidth | 819 GB/s | 273 GB/s |
| FP16 compute | 28 TFLOPS | 62 TFLOPS |
| Software | Metal | CUDA |
| Unified memory | 96 GB | 128 GB |
| GPU can use by default | 72 GB | 119 GB |
| GPU can use, tuned | 88 GB | 122 GB |
| Operating system | macos | dgx-os |
| Processor | M3 Ultra 32-core | 20-core Arm (GB10) |
| Storage | 1000 GB | 4000 GB |
What each runs
Estimated decode speed with the most precise quant that fits each one est.
| Model | Mac Studio M3 Ultra 96GB | NVIDIA DGX Spark 128GB |
|---|---|---|
| DeepSeek V4 Flash | won't run | 6tok/sUD-Q2_K_XL |
| GLM 5.3 Flash | won't run | 10tok/sUD-Q2_K_XL |
| Qwen 3.8 Flash Next | won't run | 3tok/sUD-Q4_K_XL |
| Qwen 3.8 27B | 14tok/sQ8_0 | 5tok/sQ8_0 |
| MiniMax M2.7 | won't run | 3tok/sUD-Q2_K_XL |
| Qwen 3.5 122B-A10B | 41tok/sQ3_K_M | 9tok/sQ6_K |
| Qwen 3.6 27B | 14tok/sQ8_0 | 5tok/sQ8_0 |
| Qwen3 Coder Next | 49tok/sQ6_K | 14tok/sQ8_0 |
| Nemotron 3 Super 120B-A12B | 44tok/sUD-Q3_K_M | 8tok/sUD-Q6_K |
| gpt-oss 120B | 40tok/sMXFP4 | 13tok/sMXFP4 |
| Gemma 4 31B | 12tok/sQ8_0 | 4tok/sQ8_0 |
| Mistral Small 4 119B | 51tok/sUD-Q3_K_M | 11tok/sUD-Q6_K |
| Qwen 3.6 35B-A3B | 55tok/sQ8_0 | 17tok/sQ8_0 |
| Qwen 3.5 27B | 14tok/sQ8_0 | 5tok/sQ8_0 |
| Gemma 4 26B-A4B | 45tok/sQ8_0 | 14tok/sQ8_0 |
| GLM 4.5 Air | 10tok/sQ2_K | 3tok/sQ6_K |
| Qwen3 Next 80B-A3B | 56tok/sQ6_K | 16tok/sQ8_0 |
| Qwen 3.5 35B-A3B | 55tok/sQ8_0 | 17tok/sQ8_0 |
| GLM 4.7 Flash | 26tok/sQ8_0 | 8tok/sQ8_0 |
| Devstral Small 2 24B | 11tok/sQ8_0 | 4tok/sQ8_0 |
| Nemotron 3.5 Lightning 30B-A3B | 76tok/sUD-Q6_K_XL | 24tok/sUD-Q6_K_XL |
| Kimi Linear 48B-A3B | 75tok/sQ8_0 | 24tok/sQ8_0 |
| gpt-oss 20B | 57tok/sMXFP4 | 18tok/sMXFP4 |
| Qwen3 Coder 30B-A3B | 20tok/sQ8_0 | 6tok/sQ8_0 |
| Qwen3 32B | 8tok/sQ8_0 | 2tok/sQ8_0 |
| Gemma 4 12B | 35tok/sQ8_0 | 11tok/sQ8_0 |
| Qwen 3.5 9B | 38tok/sQ8_0 | 12tok/sQ8_0 |
| Gemma 3 27B | 13tok/sQ8_0 | 4tok/sQ8_0 |
| Mistral Small 3.2 24B | 11tok/sQ8_0 | 4tok/sQ8_0 |
| DeepSeek R1 Distill Qwen 32B | 8tok/sQ8_0 | 2tok/sQ8_0 |
| Llama 3.3 70B | won't run | 1tok/sQ8_0 |
| Ministral 3 14B | 15tok/sQ8_0 | 5tok/sQ8_0 |
| Qwen3 14B | 14tok/sQ8_0 | 5tok/sQ8_0 |
| DeepSeek R1 0528 Qwen3 8B | 19tok/sQ8_0 | 6tok/sQ8_0 |
| Qwen3 8B | 19tok/sQ8_0 | 6tok/sQ8_0 |
| Phi-4 14B | 13tok/sQ8_0 | 4tok/sQ8_0 |
| Qwen 3.5 4B | 60tok/sQ8_0 | 19tok/sQ8_0 |
| Ministral 3 8B | 20tok/sQ8_0 | 6tok/sQ8_0 |
| Granite 4.2 8B | 17tok/sQ8_0 | 6tok/sQ8_0 |
| Gemma 4 E4B | 44tok/sQ8_0 | 14tok/sQ8_0 |
| LFM2.5 8B-A1B | 43tok/sQ8_0 | 14tok/sQ8_0 |
| Llama 3.1 8B | 22tok/sQ6_K | 7tok/sQ6_K |
Questions
- Which is better for local AI, the Mac Studio M3 Ultra 96GB or the NVIDIA DGX Spark 128GB?
- Choose the NVIDIA DGX Spark 128GB to run bigger models. Choose the Mac Studio M3 Ultra 96GB 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 DGX Spark 128GB?
- At 128k context the NVIDIA DGX Spark 128GB runs 42 of the models we track fully in memory and the Mac Studio M3 Ultra 96GB 37; only the NVIDIA DGX Spark 128GB holds DeepSeek V4 Flash.
- Which is faster for local AI, the Mac Studio M3 Ultra 96GB or the NVIDIA DGX Spark 128GB?
- On the models both run fully in memory, the Mac Studio M3 Ultra 96GB is about 3.2× as fast as the NVIDIA DGX Spark 128GB (estimated, median across models).
- What is the price difference between the Mac Studio M3 Ultra 96GB and the NVIDIA DGX Spark 128GB?
- In Germany the Mac Studio M3 Ultra 96GB costs 5.999 € at eBay and the NVIDIA DGX Spark 128GB 8.313 € at eBay: the Mac Studio M3 Ultra 96GB is 2.314 € cheaper.
More comparisons
- ASUS Ascent GX10 128GB vs Mac Studio M3 Ultra 96GB
- ASUS Ascent GX10 128GB vs NVIDIA DGX Spark 128GB
- Framework Desktop Ryzen AI Max+ 395 128GB vs NVIDIA DGX Spark 128GB
- Mac Studio M3 Ultra 256GB vs Mac Studio M3 Ultra 96GB
- Mac Studio M3 Ultra 512GB vs Mac Studio M3 Ultra 96GB
- Mac Studio M3 Ultra 96GB vs Mac Studio M4 Max 128GB
- Mac Studio M3 Ultra 96GB vs Mac Studio M4 Max 64GB
- Mac Studio M3 Ultra 96GB vs NVIDIA GeForce RTX 5090
- Mac Studio M3 Ultra 96GB vs NVIDIA RTX PRO 6000 Blackwell
- Mac Studio M4 Max 128GB vs NVIDIA DGX Spark 128GB