Back to search results

Qwen3.8-27B · IQ2_XS

Start a discussion
Apple M5 Ultra512GBllama.cpp v0.47L2 Cross-framework verified

This page aggregates 2 real-world runs of Qwen3.8-27B (IQ2_XS) on Apple M5 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

89.58 tok/s

Decode

Decode speed

565.47 tok/s

Prefill

Prefill speed

0.33 s

TTFT

Time to first token

19.46 GB

VRAM

VRAM usage

L2 Cross-framework verified

2

Measured runs

1

Independent sources

Other

Source platforms

2026-10-19

Last verified

Performance

  1. llama.cpp · IQ2_XS (current)Decode 89.58 · Prefill 565.47 ·
  2. Ollama · Q2_K Decode 74.38 · Prefill 283.32 ·
  3. Ollama · Q3_K_S Decode 60.73 · Prefill 250.32 ·
  4. llama.cpp · Q5_K_M Decode 46.8 · Prefill 351 ·
  5. llama.cpp · Q4_K_M Decode 57.1 · Prefill 428.3 ·

Core figures

Decode (avg)
89.58 tok/s
Prefill (avg)
565.47 tok/s
TTFT (avg)
0.33 s
VRAM (avg)
19.46 GB
MTP acceptance rate
—
TTFB
0.61 GB
Power draw
129.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.8-27B
Quantization
IQ2_XS
Framework
llama.cpp
Version
v0.47
Context length
16384 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)
19.46 GB
OS
macOS 26.1
Power draw
129.5 W

Get started

Original model resources

Base model resources may not match this quantization. Use the verified deployment weights above when available.

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.47 · macOS 26.1 · 16384 ctx

    76.26 tok/s

    Decode

    359.13 tok/s

    Prefill

    0.53 s

    TTFT

    29.32 GB

    VRAM

    —

    MTP

    129.5 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=29 running model_id=5 at IQ2_XS
  2. L0 Self-reportedOtherOriginal link Verified on 2026-10-06

    b4200 · Linux · CUDA 12.6 · 4096 ctx

    102.9 tok/s

    Decode

    771.8 tok/s

    Prefill

    0.13 s

    TTFT

    9.6 GB

    VRAM

    —

    MTP

    — W

    Power draw

    L0 理论估算(add-500-records-batch2)