Spark Bench

Nemotron 3.5 Super VL 120B-A12B NVFP4 on vLLM with DFlash, one Spark, thinking off

1× DGX Spark · vLLM ·

Tested configuration

Hardware
1x NVIDIA DGX Spark (128GB)
Serving runtime
vLLM
Runtime version or build
vLLM 0.32.0rc1.dev190+g3f1bcfd82
Weight quantization or precision
NVFP4
Benchmark date
Recipe credit
Kindling Nemotron 3.5 Super VL DGX Spark recipeBy kindlingai452acc1d74e76cfee4b45ad7a16998ee57f0a423

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)
134.150.58
234.0116.58
434.5647.22
834.17105.30
1234.95167.49
1634.85227.70

Prefill and context

Configured request window
1,048,576 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 tokens980.801.86
32,768 input tokens1,934.0714.85
131,072 input tokens1,839.1069.19
245,760 input tokensPasskey prompt1,577.68155.41

Sustained load

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

Aggregate throughput
33.90 output tok/s
TTFT p95
113.03 s
Completed requests
87
Request errors
0

Serving regression checks 13/13 passed

Tool-use evaluation

Complete · 100% completionSafety gate failed

Compare only against complete results with 100% completion under the same protocol ID.

Protocol ID
tool-eval-standard-v1-b60e2f0
Workload
69 mock-tool scenarios per trial · 3 trials
Mean score
92.7 / 100
Pass@3 · passed at least once
97.1%
Pass³ · passed every trial
81.2%

Read the tool-eval report and scenario findings