Back to search results
Qwen3.5-4B · IQ2_M
Intel Arc 140Vllama.cpp v0.21L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-4B (IQ2_M) on Intel Arc 140V with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
86.1 tok/s
Decode
Decode speed
275.57 tok/s
Prefill
Prefill speed
2.15 s
TTFT
Time to first token
6.47 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 · IQ2_M (current)Decode 86.1 · Prefill 275.57 ·
- llama.cpp · Q4_K_S Decode 66.95 · Prefill 214.51 ·
- vLLM · Q5_K_S Decode 56.56 · Prefill 269.34 ·
Core figures
- Decode (avg)
- 86.1 tok/s
- Prefill (avg)
- 275.57 tok/s
- TTFT (avg)
- 2.15 s
- VRAM (avg)
- 6.47 GB
- MTP acceptance rate
- —
- TTFB
- 0.69 GB
- Power draw
- 171.5 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-4B
- Quantization
- IQ2_M
- Framework
- llama.cpp
- Version
- v0.21
- Context length
- 8192 tokens
- Batch size
- 256
- GPU layers
- 33
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Intel Arc 140V
- Measured VRAM (avg)
- 6.47 GB
- OS
- Windows 11 23H2
- Driver
- Intel Arc 31.0.101.5186
- CUDA
- 12.6
- Power draw
- 171.5 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.21 · Windows 11 23H2 · CUDA 12.6 · 8192 ctx
86.1 tok/s
Decode
275.57 tok/s
Prefill
2.15 s
TTFT
6.47 GB
VRAM
—
MTP
171.5 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=98 running model_id=48 at IQ2_M