NVIDIA DGX Spark vs M1 Ultra 128GB for local LLMs
86 reports on the NVIDIA DGX Spark and 11 on the M1 Ultra 128GB
Which is faster for local LLMs?
On gpt-oss 117B · 5.1B active at 4-bit, the NVIDIA DGX Spark has 52.9 generation tokens/s (one run); prompt processing 1689.5 tokens/s (one run); the M1 Ultra 128GB has 65.5 generation tokens/s (one run); prompt processing 392.4 tokens/s (one run). Those are community reports whose engines and settings differ, so the gap is a rough sign, not a controlled measure of the two cards. Memory bandwidth, which caps how fast a card can write, is 3.0x higher on the M1 Ultra 128GB (273 GB/s on the NVIDIA DGX Spark, 819 on the M1 Ultra 128GB).
- Memory
- 128GB unified
- Memory bandwidth
- 273 GB/s
- FP16 compute
- 125 TFLOPS
- Reports
- 86
- Memory
- 128GB unified
- Memory bandwidth
- 819 GB/s
- FP16 compute
- 21 TFLOPS
- Reports
- 11
Measured on both
Generation and prompt processing speeds from plain runs (one device, one request, no speculative decoding, the whole model in the device's memory) of the same model size at the same bit class. Each side shows the engines and reported context behind its median. Context can mean a configured window or prompt depth, so these are community comparisons with differing settings. How to read these
| Model | NVIDIA DGX Spark | M1 Ultra 128GB |
|---|---|---|
| gpt-oss 117B · 5.1B active4-bit | 52.9one runPP 1689.5 (one run)llama.cppcontext not stated | 65.5one runPP 392.4 (one run)engine not statedcontext not stated |
Estimated by the calculator
The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape. Where a model was also measured above, the measurement wins.
| Model | NVIDIA DGX Spark | M1 Ultra 128GB |
|---|---|---|
| gpt-oss 117B · 5.1B active | 18.0 t/sQ6_K | 42.7 t/sQ5_K_S |
| Qwen3.8 27B | 4.0 t/sFP16 | 7.5 t/sFP16 |
| Qwen3.8 125B · 6B active | 14.4 t/sQ6_K | 34.0 t/sQ5_K_S |
| Qwen3.6 35B · 3B active | 14.5 t/sFP16 | 23.5 t/sFP16 |
| DeepSeek V4 Flash 284B · 13B active | 38.3 t/sQ2_K | 53.5 t/sQ2_K |
| Qwen3.6 27B | 4.0 t/sFP16 | 7.5 t/sFP16 |
| Gemma 4 26B · 4B active | 10.8 t/sFP16 | 20.2 t/sFP16 |
Change the context, RAM or model in the calculator: NVIDIA DGX Spark · M1 Ultra 128GB
Every report on each card, with its source: NVIDIA DGX Spark · M1 Ultra 128GB