Back to search results
Qwen3.5-4B · Q3_K_L
Apple M5 Max128GBllama.cpp v0.37L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-4B (Q3_K_L) on Apple M5 Max with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
104.81 tok/s
Decode
Decode speed
351.35 tok/s
Prefill
Prefill speed
2.2 s
TTFT
Time to first token
8.15 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
17 days ago
Last verified
Performance
- llama.cpp · Q3_K_L (current)Decode 104.81 · Prefill 351.35 ·
- Ollama · IQ1_S Decode 115.64 · Prefill 505.04 ·
- Ollama · Q5_K_S Decode 65.93 · Prefill 214.6 ·
- llama.cpp · IQ3_S Decode 279.98 · Prefill 1354 ·
- Strata · Q3_K_L Decode 354.31 · Prefill 1310.91 ·
- Strata · Q5_K_S Decode 279.1 · Prefill 1192.93 ·
- vLLM · IQ2_M Decode 364.07 · Prefill 1250.03 ·
- vLLM · Q4_K_S Decode 256.46 · Prefill 848.78 ·
- Strata · Q5_K_M Decode 272.64 · Prefill 963.9 ·
Core figures
- Decode (avg)
- 104.81 tok/s
- Prefill (avg)
- 351.35 tok/s
- TTFT (avg)
- 2.2 s
- VRAM (avg)
- 8.15 GB
- MTP acceptance rate
- —
- TTFB
- 0.84 GB
- Power draw
- 49 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-4B
- Quantization
- Q3_K_L
- Framework
- llama.cpp
- Version
- v0.37
- Context length
- 8192 tokens
- Batch size
- 512
- GPU layers
- 28
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Apple M5 Max
- Nominal VRAM
- 128 GB
- Measured VRAM (avg)
- 8.15 GB
- OS
- macOS 26.1
- Power draw
- 49 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.37 · macOS 26.1 · 8192 ctx
104.81 tok/s
Decode
351.35 tok/s
Prefill
2.2 s
TTFT
8.15 GB
VRAM
—
MTP
49 W
Power draw
L2 batch seed 2026-10-19: llama.cpp on hw_id=28 running model_id=48 at Q3_K_L