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.
| Concurrency | Output tok/s | TTFT p95 (s) |
|---|---|---|
| 1 | 38.43 | 0.41 |
| 2 | 68.64 | 0.82 |
| 4 | 151.70 | 1.28 |
| 8 | 215.06 | 1.58 |
| 12 | 153.55 | 29.90 |
| 16 | 184.47 | 30.48 |
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/s | TTFT p95 (s) |
|---|---|---|
| 4,096 input tokens | 1,015.76 | 2.55 |
| 32,768 input tokens | 1,551.13 | 19.36 |
| 131,072 input tokens | 1,808.72 | 83.49 |
| 245,760 input tokensPasskey prompt | 409.67 | 598.56 |
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.