Back to search results
Qwen3.5-9B · IQ2_M
NVIDIA DGX Spark128GBllama.cpp v0.49L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-9B (IQ2_M) on NVIDIA DGX Spark with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
181.94 tok/s
Decode
Decode speed
684.54 tok/s
Prefill
Prefill speed
2.21 s
TTFT
Time to first token
11.93 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 · Q6_K Decode 110.08 · Prefill 496.59 ·
- Strata · Q3_K_M Decode 209.28 · Prefill 1036.76 ·
- llama.cpp · IQ2_M (current)Decode 181.94 · Prefill 684.54 ·
- vLLM · IQ2_M Decode 210.69 · Prefill 1114.84 ·
- vLLM · Q8_0 Decode 101.73 · Prefill 523.44 ·
Core figures
- Decode (avg)
- 181.94 tok/s
- Prefill (avg)
- 684.54 tok/s
- TTFT (avg)
- 2.21 s
- VRAM (avg)
- 11.93 GB
- MTP acceptance rate
- —
- TTFB
- 1.46 GB
- Power draw
- 213.3 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-9B
- Quantization
- IQ2_M
- Framework
- llama.cpp
- Version
- v0.49
- Context length
- 8192 tokens
- Batch size
- 512
- GPU layers
- 99
- 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)
- 11.93 GB
- OS
- Ubuntu 24.04
- Driver
- 555.85
- CUDA
- 12.8
- Power draw
- 213.3 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.49 · Ubuntu 24.04 · CUDA 12.8 · 8192 ctx
181.94 tok/s
Decode
684.54 tok/s
Prefill
2.21 s
TTFT
11.93 GB
VRAM
—
MTP
213.3 W
Power draw
L2 batch seed 2026-10-13: llama.cpp on hw_id=54 running model_id=49 at IQ2_M