DiggTop

DGX Spark vs Intel Arc Pro B70

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
Intel Arc Pro B7016 verified runs · best 283.1 tok/s
ModelDGX SparkIntel Arc Pro B70
Laguna-XS-2.1283.1 Q5_K_M GGUF
Qwen3.6-35B-A3B248.3 GPTQ-4bit
Ornith-1.0-35B-B70-Turbo224.9 Q5_K_M GGUF
Nex-N2-mini-B70-Turbo224.3 Q5_K_M GGUF
SIQ-1-35B-B70-Turbo223.7 Q5_K_M GGUF
Qwen-AgentWorld-35B-A3B-B70-Turbo209.0 Q5_K_M GGUF
NVIDIA-Nemotron-3.5-Lightning-30B-A3B186.6 GPTQ-INT4-G64-sym-local+DFlash-B
LFM2.5-2.6B132.4 Q8_0
Ornith-1.5-35B-A3B131.5 Q4_K_M
Qwen 3.8 Flash86.8 EXL3 3.05bpw
Qwen3.8-Flash-Next72.0 EXL3
GLM-5.3-Flash64.1 EXL3 2.05 bpw
Qwen3.8-Flash-Next-hibrid4860.0 NVFP4 (output head)
Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP58.4 Q4_K_M
Qwen3.8-27B56.0 NVFP4 W4A427.8 Q4_K_M
Ornith-1.5-9B49.6 Q8_0
DeepSeek-V4.1-Flash47.0 FP4
Qwen3.6-27B-MTP36.0 Q8_0
Muse-Glimmer-30B29.2 Q4_K_XL
ThinkingCap-Qwen3.6-27B27.9 Q6_K

Head-to-head (models tested on both): DGX Spark 1 · Intel Arc Pro B70 0 · tied 0 · data as of Sep 23, 2026. Full rules: methodology.

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