Arao AI
Back to search results

Qwen3-Coder-Next · Q5_K_S

Start a discussion
Apple M5 Max128GBllama.cpp v0.39L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-Coder-Next (Q5_K_S) on Apple M5 Max with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

144.1 tok/s

Decode

Decode speed

764.74 tok/s

Prefill

Prefill speed

1.83 s

TTFT

Time to first token

1.05 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. llama.cpp · Q5_K_S (current)Decode 144.1 · Prefill 764.74 ·
  2. Strata · Q2_K Decode 312.25 · Prefill 1150.01 ·

Core figures

Decode (avg)
144.1 tok/s
Prefill (avg)
764.74 tok/s
TTFT (avg)
1.83 s
VRAM (avg)
1.05 GB
MTP acceptance rate
—
TTFB
1.15 GB
Power draw
261.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-Coder-Next
Quantization
Q5_K_S
Framework
llama.cpp
Version
v0.39
Context length
8192 tokens
Batch size
256
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)
1.05 GB
OS
Ubuntu 22.04
Driver
565.57
CUDA
12.8
Power draw
261.4 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-16

    v0.39 · Ubuntu 22.04 · CUDA 12.8 · 8192 ctx

    144.1 tok/s

    Decode

    764.74 tok/s

    Prefill

    1.83 s

    TTFT

    1.05 GB

    VRAM

    —

    MTP

    261.4 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=28 running model_id=58 at Q5_K_S