DGX Spark vs H200
For running local LLMs · 26 reports across 10 models
Highest reported generation speed by model family
These maxima come from different community setups. Model size, quantization, GPU count, offloading and concurrency may differ. They do not establish which GPU is faster for the same workload. Check the comparison conditions →
| Model | DGX Spark | H200 |
|---|---|---|
| DeepSeek V4 Flashup to 285B | 90.0n=10 | 4.8n=2 |
| Ling-3.0up to 124B | 40.9n=3 | — |
| Qwen3.8up to 2400B | 181.0n=3 | — |
| GLM-5.2up to 744B | 24.0n=2 | — |
| DeepSeek V3 | —n=1 | — |
| GLM-5.1 | —n=1 | — |
| HobbyLMup to 1B | — | —n=1 |
| Kimi K3 | 20.0n=1 | — |
| Museup to 30B | 37.0n=1 | — |
| Qwen3.6up to 35B | —n=1 | — |