Back to search results
Qwen3-8B · Q4_K_S
Apple M2 Ultrallama.cpp v0.43L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-8B (Q4_K_S) on Apple M2 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
108.67 tok/s
Decode
Decode speed
436.07 tok/s
Prefill
Prefill speed
2.13 s
TTFT
Time to first token
15.2 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 · Q4_K_S (current)Decode 108.67 · Prefill 436.07 ·
- llama.cpp · Q5_K_S Decode 87.67 · Prefill 476.02 ·
- Ollama · Q5_K_M Decode 79.98 · Prefill 403.29 ·
- Ollama · Q8_0 Decode 67.83 · Prefill 284.74 ·
Core figures
- Decode (avg)
- 108.67 tok/s
- Prefill (avg)
- 436.07 tok/s
- TTFT (avg)
- 2.13 s
- VRAM (avg)
- 15.2 GB
- MTP acceptance rate
- —
- TTFB
- 0.87 GB
- Power draw
- 47.1 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-8B
- Quantization
- Q4_K_S
- Framework
- llama.cpp
- Version
- v0.43
- Context length
- 8192 tokens
- Batch size
- 256
- GPU layers
- 33
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Apple M2 Ultra
- Measured VRAM (avg)
- 15.2 GB
- OS
- macOS 15.0
- Driver
- Apple M2 Ultra
- CUDA
- N/A
- Power draw
- 47.1 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.43 · macOS 15.0 · CUDA N/A · 8192 ctx
108.67 tok/s
Decode
436.07 tok/s
Prefill
2.13 s
TTFT
15.2 GB
VRAM
—
MTP
47.1 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=103 running model_id=53 at Q4_K_S