Qwen3.8-27B · Q4_K_M
This page aggregates 1 real-world runs of Qwen3.8-27B (Q4_K_M) on Apple M5 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
57.1 tok/s
Decode
Decode speed
428.3 tok/s
Prefill
Prefill speed
0.23 s
TTFT
Time to first token
16.9 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- llama.cpp · IQ2_XS Decode 102.9 · Prefill 771.8 ·
- llama.cpp · Q5_K_M Decode 46.8 · Prefill 351 ·
- llama.cpp · Q4_K_M (current)Decode 57.1 · Prefill 428.3 ·
Core figures
- Decode (avg)
- 57.1 tok/s
- Prefill (avg)
- 428.3 tok/s
- TTFT (avg)
- 0.23 s
- VRAM (avg)
- 16.9 GB
- MTP acceptance rate
- —
- TTFB
- — GB
- Power draw
- — W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.8-27B
- Quantization
- Q4_K_M
- Framework
- llama.cpp
- Version
- b4200
- Context length
- 4096 tokens
- Batch size
- 512
- 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 M5 Ultra
- Nominal VRAM
- 512 GB
- Measured VRAM (avg)
- 16.9 GB
- OS
- Linux
- Driver
- 570.65
- CUDA
- 12.6
Get started
Original model resources
Base model resources may not match this quantization. Use the verified deployment weights above when available.
Sources & evidence
1 measured records in total, each traceable to its original source
b4200 · Linux · CUDA 12.6 · 4096 ctx
57.1 tok/s
Decode
428.3 tok/s
Prefill
0.23 s
TTFT
16.9 GB
VRAM
—
MTP
— W
Power draw
L0 理论估算（add-500-records-batch2）