Spark Bench

Qwen3.8-Flash-Next-NVFP4 MiaAI vLLM two-Spark 2200 MHz baseline

2× DGX Spark · vLLM ·

Tested configuration

Hardware
2x NVIDIA DGX Spark (128GB each)
Serving runtime
vLLM
Runtime version or build
vLLM 0.1.dev20073+g8e685d198, MTP speculative decoding with 3 draft tokens · vllm/vllm-openai:qwen38-flash-next
Weight quantization or precision
NVFP4
Benchmark date
Recipe credit
MiaAI-Lab Qwen dual-Spark recipeBy MiaAI-Lab169fbad266f2791335a3102f0d3d625e7c295563

Decode sweep

~256 input / 512 output tokens per request. Output throughput is aggregate across concurrent requests. TTFT means time to first token.

ConcurrencyOutput tok/sTTFT p95 (s)
147.510.36
281.830.83
4135.451.25
8213.721.18
12210.1919.07
16210.0320.76

Prefill and context

Configured request window
1,000,000 tokens
Longest tested prompt
245,760 target input tokens
Validated retrieval
≈240K input tokens · passkey passed

Input throughput uses full request wall time. The passkey measures retrieval from one long prompt.

Workload (c1)Input tok/sTTFT p95 (s)
4,096 input tokens1,251.711.96
32,768 input tokens2,361.8512.64
131,072 input tokens2,350.8754.38
245,760 input tokensPasskey prompt2,146.46113.18

Sustained load

20-minute soak at concurrency 8 · ~1024 input / 512 output tokens per request.

Aggregate throughput
201.58 output tok/s
TTFT p95
1.31 s
Completed requests
477
Request errors
0

Serving regression checks 13/13 passed

Tool-use evaluation

Unavailable · no published tool-eval result for this configuration.

Other configurations for Qwen3.8-Flash-Next NVFP4