Back to search results
Qwen3-4B · Q8_0
Apple M1llama.cpp v0.44L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-4B (Q8_0) on Apple M1 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
36.81 tok/s
Decode
Decode speed
167.78 tok/s
Prefill
Prefill speed
0.92 s
TTFT
Time to first token
11.76 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 · Q3_K_L Decode 65.56 · Prefill 344.27 ·
- llama.cpp · Q8_0 (current)Decode 36.81 · Prefill 167.78 ·
- Ollama · Q2_K Decode 72.5 · Prefill 393.97 ·
Core figures
- Decode (avg)
- 36.81 tok/s
- Prefill (avg)
- 167.78 tok/s
- TTFT (avg)
- 0.92 s
- VRAM (avg)
- 11.76 GB
- MTP acceptance rate
- —
- TTFB
- 0.29 GB
- Power draw
- 21.9 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-4B
- Quantization
- Q8_0
- Framework
- llama.cpp
- Version
- v0.44
- Context length
- 16384 tokens
- Batch size
- 256
- GPU layers
- 99
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Apple M1
- Measured VRAM (avg)
- 11.76 GB
- OS
- macOS 15.0
- Driver
- Apple M1
- CUDA
- N/A
- Power draw
- 21.9 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.44 · macOS 15.0 · CUDA N/A · 16384 ctx
36.81 tok/s
Decode
167.78 tok/s
Prefill
0.92 s
TTFT
11.76 GB
VRAM
—
MTP
21.9 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=99 running model_id=54 at Q8_0