Spark Bench

Qwen3.8-Flash-Next-NVFP4 serving baseline

2× DGX Spark · SGLang ·

Tested configuration

Hardware
2x NVIDIA DGX Spark (128GB each)
Serving runtime
SGLang
Runtime version or build
Unavailable
Weight quantization or precision
NVFP4
Benchmark date
Recipe credit
Qwen3.8 Flash Next DGX Spark recipeBy tomrhudson

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)
138.430.41
268.640.82
4151.701.28
8215.061.58
12153.5529.90
16184.4730.48

Prefill and context

Configured request window
262,144 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 tokens1,015.762.55
32,768 input tokens1,551.1319.36
131,072 input tokens1,808.7283.49
245,760 input tokensPasskey prompt409.67598.56

Sustained load

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

Aggregate throughput
157.61 output tok/s
TTFT p95
1.64 s
Completed requests
374
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