Back to search results
Qwen3.5-4B · Q4_K_M
Qualcomm Snapdragon 7 Gen 4llama.cpp v0.54L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-4B (Q4_K_M) on Qualcomm Snapdragon 7 Gen 4 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
50.36 tok/s
Decode
Decode speed
218.21 tok/s
Prefill
Prefill speed
1.39 s
TTFT
Time to first token
7.6 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- llama.cpp · Q4_K_M (current)Decode 50.36 · Prefill 218.21 ·
- Ollama · IQ4_XS Decode 48.35 · Prefill 193.59 ·
Core figures
- Decode (avg)
- 50.36 tok/s
- Prefill (avg)
- 218.21 tok/s
- TTFT (avg)
- 1.39 s
- VRAM (avg)
- 7.6 GB
- MTP acceptance rate
- —
- TTFB
- 0.4 GB
- Power draw
- 21.3 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-4B
- Quantization
- Q4_K_M
- Framework
- llama.cpp
- Version
- v0.54
- Context length
- 4096 tokens
- Batch size
- 256
- GPU layers
- 20
- Flash Attention
- Off
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Qualcomm Snapdragon 7 Gen 4
- Measured VRAM (avg)
- 7.6 GB
- OS
- Windows 11 23H2
- Driver
- Qualcomm Adreno 7.0
- CUDA
- N/A
- Power draw
- 21.3 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.54 · Windows 11 23H2 · CUDA N/A · 4096 ctx
50.36 tok/s
Decode
218.21 tok/s
Prefill
1.39 s
TTFT
7.6 GB
VRAM
—
MTP
21.3 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=105 running model_id=48 at Q4_K_M