Arao AI
Back to search results

Qwen3-1.7B · Q2_K

Start a discussion
Apple M5 Max128GBvLLM v0.48L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-1.7B (Q2_K) on Apple M5 Max with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

302.27 tok/s

Decode

Decode speed

1,354.5 tok/s

Prefill

Prefill speed

0.36 s

TTFT

Time to first token

4.06 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. vLLM · Q2_K (current)Decode 302.27 · Prefill 1354.5 ·
  2. vLLM · Q8_0 Decode 255.36 · Prefill 1217.53 ·
  3. Strata · Q8_0 Decode 319.85 · Prefill 1662.65 ·
  4. Ollama · Q3_K_L Decode 353.08 · Prefill 1662.13 ·

Core figures

Decode (avg)
302.27 tok/s
Prefill (avg)
1,354.5 tok/s
TTFT (avg)
0.36 s
VRAM (avg)
4.06 GB
MTP acceptance rate
—
TTFB
2.42 GB
Power draw
224 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-1.7B
Quantization
Q2_K
Framework
vLLM
Version
v0.48
Context length
32768 tokens
Batch size
8
GPU layers
999
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)
4.06 GB
OS
Windows 11 23H2
Driver
565.57
CUDA
12.6
Power draw
224 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

    v0.48 · Windows 11 23H2 · CUDA 12.6 · 32768 ctx

    302.27 tok/s

    Decode

    1,354.5 tok/s

    Prefill

    0.36 s

    TTFT

    4.06 GB

    VRAM

    —

    MTP

    224 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=28 running model_id=55 at Q2_K