Different platforms, same budget
NVIDIA DGX Spark vs Apple Mac Studio · M3 Ultra
Memory, bandwidth, software stack, price and the models each one can hold. Generated from the two sourced machine records; estimates are labelled.
NVIDIA GB10
NVIDIA DGX Spark
128 GB unified
Open the record
Apple Silicon
Apple Mac Studio · M3 Ultra
96 GB unified · reference
Open the recordWhat actually differs
- Different software stacks: the NVIDIA DGX Spark runs CUDA, the Apple Mac Studio · M3 Ultra runs MLX and Metal. Check that your runtime and model format are supported before comparing numbers.
- The NVIDIA DGX Spark has 32 GB more model memory (128 GB vs 96 GB). That decides which models fit at all.
- The Apple Mac Studio · M3 Ultra reads memory 3.0× faster (819 vs 273 GB/s). For a model that fits on both, that is the ceiling on generation speed.
- No public list price in the record for the Apple Mac Studio · M3 Ultra. Ask the supplier for a configured quote.
- Form factor: Desktop AI system vs Mini workstation.
Which models fit
Estimate · 4-bit · 8K context · 1 request| Model | NVIDIA DGX Spark | Apple Mac Studio · M3 Ultra |
|---|---|---|
| Llama 3.1 8B | Fits 16 GB≤ 48 tok/s | Fits 16 GB≤ 144 tok/s |
| Qwen3 14B | Fits 20 GB≤ 28 tok/s | Fits 20 GB≤ 83 tok/s |
| Qwen3 32B | Fits 32 GB≤ 13 tok/s | Fits 32 GB≤ 39 tok/s |
| Llama 3.3 70B | Fits 54 GB≤ 6.3 tok/s | Fits 54 GB≤ 19 tok/s |
| gpt-oss-20b (MoE) | Fits 23 GB≤ 110 tok/s | Fits 23 GB≤ 331 tok/s |
| gpt-oss-120b (MoE) | Fits 80 GB≤ 77 tok/s | Fits 80 GB≤ 232 tok/s |
| Qwen3 235B-A22B (MoE) | Too large 151 GB | Too large 151 GB |
| DeepSeek V3 / R1 671B (MoE, MLA) | Too large 408 GB | Too large 408 GB |
Specifications
Sourced and dated| NVIDIA DGX Spark | Apple Mac Studio · M3 Ultra | |
|---|---|---|
| Chip / accelerator | GB10 Grace Blackwell Superchip | Apple M3 Ultra |
| Platform | NVIDIA GB10 | Apple Silicon |
| Model memory | 128 GB unified | 96 GB unified · reference |
| Memory bandwidth | 273 GB/s | 819 GB/s |
| Software stack | CUDA | MLX and Metal |
| Operating system | NVIDIA DGX OS | macOS |
| Storage | 4 TB NVMe | 1 TB · reference |
| Networking | 10 GbE · ConnectX-7 · Wi-Fi 7 | Thunderbolt 5 · 10 GbE |
| Form factor | Desktop AI system | Mini workstation |
| Dimensions | 150 × 150 × 50.5 mm | 197 × 197 × 95 mm |
| Weight | 1.2 kg | Not verified |
| Power rating (not consumption) | 240 W adapter · 140 W SoC TDP | 480 W maximum continuous power rating · not typical consumption |
| Supplier list price | ≈ $4.7k · Supplier list price · United States · Sept 2026 | Price from the manufacturer · No public list price · manufacturer page |
| Availability | Listed for order · recheck with supplier | Earlier generation · check reseller stock |
| Model configurations in the Lab | 6 | 3 |
| Setups in the Lab | 5 | 1 |
| Record reviewed | 2026-09-13 | 2026-09-07 |
Before you choose
NVIDIA DGX Spark
- Memory capacity does not establish interactive speed or simultaneous request capacity.
- UT has documented a Spark deployment. Confirm exact software versions and support scope.
Apple Mac Studio · M3 Ultra
- This is a specific M3 Ultra configuration, not a claim that it is the latest Mac Studio.
- CUDA workloads need a different platform. UT’s complete managed stack is not validated here.
NVIDIA DGX Spark compared with
ASUS Ascent GX10 Dell Pro Max with GB10 HP ZGX Nano G1n Lenovo ThinkStation PGX Acer Veriton GN100 GIGABYTE AI TOP ATOM MSI EdgeXpert Apple Mac Studio · M5 Ultra Framework Desktop · Ryzen AI Max+ MINISFORUM MS-S1 MAX HP Z2 Mini G1a AMD Ryzen AI Halo NVIDIA DGX StationApple Mac Studio · M3 Ultra compared with
Framework Desktop · Ryzen AI Max+ Apple Mac Studio · M5 UltraOwn one of these? A measured run (model revision, runtime, context, tokens per second) is worth more than this whole page. Publish it.