Back to search results
Qwen3-4B-Thinking-2507 · IQ3_S
Apple M2 ProvLLM v0.25L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (IQ3_S) on Apple M2 Pro with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
112.31 tok/s
Decode
Decode speed
603.83 tok/s
Prefill
Prefill speed
1.05 s
TTFT
Time to first token
7.64 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- vLLM · IQ3_S (current)Decode 112.31 · Prefill 603.83 ·
- Strata · IQ2_M Decode 171.6 · Prefill 574.16 ·
- Strata · Q4_K_S Decode 119.23 · Prefill 515.15 ·
Core figures
- Decode (avg)
- 112.31 tok/s
- Prefill (avg)
- 603.83 tok/s
- TTFT (avg)
- 1.05 s
- VRAM (avg)
- 7.64 GB
- MTP acceptance rate
- —
- TTFB
- 0.9 GB
- Power draw
- 20.6 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-4B-Thinking-2507
- Quantization
- IQ3_S
- Framework
- vLLM
- Version
- v0.25
- Context length
- 32768 tokens
- Batch size
- 16
- GPU layers
- 999
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Apple M2 Pro
- Measured VRAM (avg)
- 7.64 GB
- OS
- macOS 15.0
- Driver
- Apple M2 Pro
- CUDA
- N/A
- Power draw
- 20.6 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.25 · macOS 15.0 · CUDA N/A · 32768 ctx
112.31 tok/s
Decode
603.83 tok/s
Prefill
1.05 s
TTFT
7.64 GB
VRAM
—
MTP
20.6 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=102 running model_id=57 at IQ3_S