Back to search results
Qwen3-1.7B · Q5_K_M
NVIDIA DGX Spark128GBvLLM v0.54L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-1.7B (Q5_K_M) on NVIDIA DGX Spark with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
240.21 tok/s
Decode
Decode speed
1,012.49 tok/s
Prefill
Prefill speed
2.34 s
TTFT
Time to first token
5.42 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- vLLM · Q2_K Decode 374.76 · Prefill 1672.78 ·
- vLLM · Q5_K_M (current)Decode 240.21 · Prefill 1012.49 ·
- Strata · IQ3_XS Decode 315.95 · Prefill 1231.57 ·
- Strata · Q4_K_S Decode 375.7 · Prefill 1600.4 ·
- llama.cpp · Q6_K Decode 210 · Prefill 1074.66 ·
- vLLM · Q3_K_M Decode 331.74 · Prefill 1116.84 ·
Core figures
- Decode (avg)
- 240.21 tok/s
- Prefill (avg)
- 1,012.49 tok/s
- TTFT (avg)
- 2.34 s
- VRAM (avg)
- 5.42 GB
- MTP acceptance rate
- —
- TTFB
- 1.92 GB
- Power draw
- 209.2 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-1.7B
- Quantization
- Q5_K_M
- Framework
- vLLM
- Version
- v0.54
- Context length
- 4096 tokens
- Batch size
- 8
- GPU layers
- 999
- Flash Attention
- On
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)
- 5.42 GB
- OS
- macOS 15.0
- Driver
- 565.57
- CUDA
- 12.8
- Power draw
- 209.2 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.54 · macOS 15.0 · CUDA 12.8 · 4096 ctx
240.21 tok/s
Decode
1,012.49 tok/s
Prefill
2.34 s
TTFT
5.42 GB
VRAM
—
MTP
209.2 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=54 running model_id=55 at Q5_K_M