Back to search results
Qwen3-4B · FP16
Apple M2 ProvLLM v0.44L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-4B (FP16) on Apple M2 Pro with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
48.46 tok/s
Decode
Decode speed
199.81 tok/s
Prefill
Prefill speed
1.81 s
TTFT
Time to first token
17.64 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- vLLM · FP16 (current)Decode 48.46 · Prefill 199.81 ·
- vLLM · Q6_K Decode 72.96 · Prefill 295.88 ·
- Strata · FP16 Decode 65.24 · Prefill 265.48 ·
Core figures
- Decode (avg)
- 48.46 tok/s
- Prefill (avg)
- 199.81 tok/s
- TTFT (avg)
- 1.81 s
- VRAM (avg)
- 17.64 GB
- MTP acceptance rate
- —
- TTFB
- 0.39 GB
- Power draw
- 30.3 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-4B
- Quantization
- FP16
- Framework
- vLLM
- Version
- v0.44
- 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)
- 17.64 GB
- OS
- macOS 15.0
- Driver
- Apple M2 Pro
- CUDA
- N/A
- Power draw
- 30.3 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.44 · macOS 15.0 · CUDA N/A · 32768 ctx
48.46 tok/s
Decode
199.81 tok/s
Prefill
1.81 s
TTFT
17.64 GB
VRAM
—
MTP
30.3 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=102 running model_id=54 at FP16