DGX Spark vs Instinct MI100 x4
Best verified output tok/s per model (single-stream generation). Hover a value for the quant.
DGX Spark44 verified runs · best 86.8 tok/s
Instinct MI100 x41 verified run · best 132.4 tok/s
| Model | DGX Spark | Instinct MI100 x4 |
|---|
| Qwen3.8-27B | 56.0 NVFP4 W4A4 | 132.4 INT8 |
| Qwen 3.8 Flash | 86.8 EXL3 3.05bpw | — |
| Qwen3.8-Flash-Next | 72.0 EXL3 | — |
| GLM-5.3-Flash | 64.1 EXL3 2.05 bpw | — |
| Qwen3.8-Flash-Next-hibrid48 | 60.0 NVFP4 (output head) | — |
| DeepSeek-V4.1-Flash | 47.0 FP4 | — |
Head-to-head (models tested on both): DGX Spark 0 · Instinct MI100 x4 1 · tied 0 · data as of Sep 23, 2026. Full rules: methodology.
Related reading: RTX 5090 vs 3090 · quantization · Value index · Full board