Back to search results
Qwen3-8B · Q3_K_M
NVIDIA DGX Spark128GBvLLM v0.45L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-8B (Q3_K_M) on NVIDIA DGX Spark with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
167.2 tok/s
Decode
Decode speed
861.79 tok/s
Prefill
Prefill speed
1.29 s
TTFT
Time to first token
13.52 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- vLLM · Q3_K_M (current)Decode 167.2 · Prefill 861.79 ·
- Strata · Q5_K_S Decode 197.21 · Prefill 856.13 ·
- llama.cpp · Q4_K_M Decode 146.4 · Prefill 699.64 ·
- vLLM · IQ4_XS Decode 168.28 · Prefill 770.17 ·
Core figures
- Decode (avg)
- 167.2 tok/s
- Prefill (avg)
- 861.79 tok/s
- TTFT (avg)
- 1.29 s
- VRAM (avg)
- 13.52 GB
- MTP acceptance rate
- —
- TTFB
- 1.34 GB
- Power draw
- 251.9 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-8B
- Quantization
- Q3_K_M
- Framework
- vLLM
- Version
- v0.45
- Context length
- 4096 tokens
- Batch size
- 8
- 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)
- 13.52 GB
- OS
- macOS 15.0
- Driver
- 570.86
- CUDA
- 12.8
- Power draw
- 251.9 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.45 · macOS 15.0 · CUDA 12.8 · 4096 ctx
167.2 tok/s
Decode
861.79 tok/s
Prefill
1.29 s
TTFT
13.52 GB
VRAM
—
MTP
251.9 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=54 running model_id=53 at Q3_K_M