Back to search results
Qwen2.5-Omni-3B · Q3_K_M
NVIDIA DGX Spark128GBvLLM v0.50L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen2.5-Omni-3B (Q3_K_M) on NVIDIA DGX Spark with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
258.63 tok/s
Decode
Decode speed
1,322.45 tok/s
Prefill
Prefill speed
0.49 s
TTFT
Time to first token
6.38 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 258.63 · Prefill 1322.45 ·
- vLLM · Q5_K_M Decode 183.75 · Prefill 675.77 ·
- Ollama · IQ2_M Decode 292.67 · Prefill 1302.54 ·
- Ollama · IQ4_XS Decode 218.18 · Prefill 903.32 ·
Core figures
- Decode (avg)
- 258.63 tok/s
- Prefill (avg)
- 1,322.45 tok/s
- TTFT (avg)
- 0.49 s
- VRAM (avg)
- 6.38 GB
- MTP acceptance rate
- —
- TTFB
- 2.07 GB
- Power draw
- 260.4 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen2.5-Omni-3B
- Quantization
- Q3_K_M
- Framework
- vLLM
- Version
- v0.50
- Context length
- 4096 tokens
- Batch size
- 32
- 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)
- 6.38 GB
- OS
- macOS 15.0
- Driver
- 555.85
- CUDA
- 12.6
- Power draw
- 260.4 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.50 · macOS 15.0 · CUDA 12.6 · 4096 ctx
258.63 tok/s
Decode
1,322.45 tok/s
Prefill
0.49 s
TTFT
6.38 GB
VRAM
—
MTP
260.4 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=54 running model_id=60 at Q3_K_M