DGX Spark vs RTX PRO 6000 Blackwell x2

Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.

DGX Spark52 sourced runs · best 86.8 tok/s

RTX PRO 6000 Blackwell x23 verified runs · best 179.5 tok/s

Specs and evidence

	DGX Spark	RTX PRO 6000 Blackwell x2
VRAM	128 GB	192 GB
Sourced runs on the board	52	3
Models measured	9	2
Best output tok/s	86.8 tok/s	179.5 tok/s
Highest measured peak VRAM	—	186.6 GB
Street price (latest snapshot)	—	—

Sourced rows only. Peak VRAM is recorded by the person who ran the benchmark, so it is a floor for what the card can hold, not a specification. Price snapshots.

Head-to-head

Only one model has been measured on both cards (GLM-5.3-Flash). On that single data point DGX Spark records 64.1 tok/s and RTX PRO 6000 Blackwell x2 records 159.3 tok/s — treat it as one data point, not a ranking.

The measured matrix

Model	DGX Spark	RTX PRO 6000 Blackwell x2
DeepSeek-V4-Flash-Vision-Exp	—	179.5 FP8
GLM-5.3-Flash	64.1 EXL3 2.05 bpw	159.3 EXL3-3.0bpw
Qwen 3.8 Flash	86.8 EXL3 3.05bpw	—
Qwen3.8-Flash-Next	75.7 4.05bpw_h6_ng6	—
CYBER-FROST-3.8	60.7 EXL3	—
Qwen3.8-Flash-Next-hibrid48	60.0 NVFP4 (output head)	—
Qwen3.8-27B	56.0 NVFP4 W4A4	—
NVIDIA-Nemotron-3.5-Super-VL-120B-A12B	48.0 NVFP4	—
DeepSeek-V4.1-Flash	47.0 FP4	—
Underdog-Saluki-27B	22.6 IQ2-mix	—

Head-to-head (models tested on both): DGX Spark 0 · RTX PRO 6000 Blackwell x2 1 · tied 0 · data as of Oct 30, 2026. Full rules: methodology.

Related reading: RTX 5090 vs 3090 · quantization · Value index · Full board