Qwen3-4B-Thinking-2507 · IQ3_XS
This page aggregates 2 real-world runs of Qwen3-4B-Thinking-2507 (IQ3_XS) on Intel Arc A770 16GB with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.
72.87 tok/s
Decode
Decode speed
292.61 tok/s
Prefill
Prefill speed
0.75 s
TTFT
Time to first token
7.31 GB
VRAM
VRAM usage
2
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- Ollama · IQ3_XS (current)Decode 72.87 · Prefill 292.61 ·
- Ollama · Q8_0 Decode 38.64 · Prefill 189.17 ·
- Strata · IQ3_S Decode 110.46 · Prefill 356.07 ·
Core figures
- Decode (avg)
- 72.87 tok/s
- Prefill (avg)
- 292.61 tok/s
- TTFT (avg)
- 0.75 s
- VRAM (avg)
- 7.31 GB
- MTP acceptance rate
- —
- TTFB
- 0.59 GB
- Power draw
- 193.6 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-4B-Thinking-2507
- Quantization
- IQ3_XS
- Framework
- Ollama
- Version
- v0.52
- Context length
- 32768 tokens
- Batch size
- 256
- GPU layers
- 20
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Intel Arc A770 16GB
- Nominal VRAM
- 16 GB
- Measured VRAM (avg)
- 7.31 GB
- OS
- Windows 11 23H2
- Driver
- Intel Arc 31.0.101.5186
- CUDA
- 12.6
- Power draw
- 157.2 W
Sources & evidence
2 measured records in total, each traceable to its original source
v0.52 · Windows 11 23H2 · CUDA 12.6 · 32768 ctx
71.16 tok/s
Decode
261.38 tok/s
Prefill
0.79 s
TTFT
7.31 GB
VRAM
—
MTP
157.2 W
Power draw
L2 batch seed 2026-10-16: Ollama on hw_id=97 running model_id=57 at IQ3_XS
v0.29 · Ubuntu 24.04 · CUDA 12.6 · 4096 ctx
74.57 tok/s
Decode
323.83 tok/s
Prefill
0.7 s
TTFT
7.31 GB
VRAM
—
MTP
230 W
Power draw
L2 batch seed 2026-10-13: Ollama on hw_id=97 running model_id=57 at IQ3_XS