DGX Spark vs RTX 3090
Best verified output tok/s per model (single-stream generation). Hover a value for the quant.
DGX Spark29 verified runs · best 64.1 tok/s
RTX 309012 verified runs · best 709.5 tok/s
| Model | DGX Spark | RTX 3090 |
|---|---|---|
| LFM2.5-1.2B-Instruct-GGUF | — | 709.5 |
| Qwen3.5-0.8B-MTP-GGUF | — | 476.1 |
| NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF | — | 353.7 |
| LFM2.5-2.6B-GGUF | — | 305.6 |
| Qwen3.5-4B-MTP-GGUF | — | 240.4 |
| Qwen3.8-27B-EXL3-3.5bpw | — | 164.6 |
| Qwen3.5-9B-MTP-GGUF | — | 164.3 |
| Qwen3.8-27B | 56.0 | 152.4 |
| Qwen/Qwen3.8-27B | — | 136.8 |
| Qwen3.8-27B-GGUF | — | 81.9 |
| GLM-5.3-Flash | 64.1 | — |
| Qwen3.8-Flash-Next-hibrid48 | 60.0 | — |
| Qwen3.8-Flash-Next-NVFP4 | 55.0 | — |
| DeepSeek-V4.1-Flash | 47.0 | — |
| Ornstein3.6-27B-MTP-NSC-ACE-SABER-GGUF | — | 32.9 |