Back to search results
Qwen3-4B · Q3_K_L
NVIDIA DGX Spark128GBvLLM v0.48L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-4B (Q3_K_L) on NVIDIA DGX Spark with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
202.71 tok/s
Decode
Decode speed
852.86 tok/s
Prefill
Prefill speed
0.55 s
TTFT
Time to first token
8.15 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
5 days ago
Last verified
Performance
- llama.cpp · Q3_K_M Decode 206.5 · Prefill 1076.24 ·
- Strata · Q3_K_S Decode 338.04 · Prefill 1544.38 ·
- Strata · Q5_K_M Decode 221.01 · Prefill 864.66 ·
- vLLM · Q3_K_L (current)Decode 202.71 · Prefill 852.86 ·
- Strata · IQ1_S Decode 323.83 · Prefill 1439.4 ·
Core figures
- Decode (avg)
- 202.71 tok/s
- Prefill (avg)
- 852.86 tok/s
- TTFT (avg)
- 0.55 s
- VRAM (avg)
- 8.15 GB
- MTP acceptance rate
- —
- TTFB
- 1.62 GB
- Power draw
- 233.2 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-4B
- Quantization
- Q3_K_L
- Framework
- vLLM
- Version
- v0.48
- Context length
- 4096 tokens
- Batch size
- 16
- GPU layers
- 999
- Flash Attention
- Off
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA DGX Spark
- Nominal VRAM
- 128 GB
- Measured VRAM (avg)
- 8.15 GB
- OS
- Ubuntu 24.04
- Driver
- 570.86
- CUDA
- 12.8
- Power draw
- 233.2 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.48 · Ubuntu 24.04 · CUDA 12.8 · 4096 ctx
202.71 tok/s
Decode
852.86 tok/s
Prefill
0.55 s
TTFT
8.15 GB
VRAM
—
MTP
233.2 W
Power draw
L2 batch seed 2026-10-13: vLLM on hw_id=54 running model_id=54 at Q3_K_L