Qwen3.8-27B · Q4_K_M
This page aggregates 1 real-world runs of Qwen3.8-27B (Q4_K_M) on Intel Arc Pro B70 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
35.6 tok/s
Decode
Decode speed
562 tok/s
Prefill
Prefill speed
— s
TTFT
Time to first token
— GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
GitHub
Source platforms
Yesterday
Last verified
Performance
- llama.cpp · Q4_K_M (current)Decode 35.6 · Prefill 562 ·
Core figures
- Decode (avg)
- 35.6 tok/s
- Prefill (avg)
- 562 tok/s
- TTFT (avg)
- — s
- VRAM (avg)
- — 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
- b10883
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Intel Arc Pro B70
- Nominal VRAM
- 32 GB
- OS
- Ubuntu 26.04
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
b10883 · Ubuntu 26.04
35.6 tok/s
Decode
562 tok/s
Prefill
— s
TTFT
0 GB
VRAM
—
MTP
— W
Power draw
| ~1k | 35.6 t/s | 33.4 t/s | 38.5 t/s | 1.8 s | 1.8 s | ... Prefill also degrades on Vulkan only: effective prefill falls 562 → 101 t/s (1k→64k)