Arao AI
Back to search results

Qwen3-4B · IQ2_M

Start a discussion
Apple M1 MaxOllama v0.53L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B (IQ2_M) on Apple M1 Max with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.

126.39 tok/s

Decode

Decode speed

412.92 tok/s

Prefill

Prefill speed

1.07 s

TTFT

Time to first token

6.47 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 · Q2_K Decode 145.16 · Prefill 623.88 ·
  2. Ollama · IQ2_M (current)Decode 126.39 · Prefill 412.92 ·
  3. Ollama · Q8_0 Decode 64.73 · Prefill 325.03 ·

Core figures

Decode (avg)
126.39 tok/s
Prefill (avg)
412.92 tok/s
TTFT (avg)
1.07 s
VRAM (avg)
6.47 GB
MTP acceptance rate
—
TTFB
1.01 GB
Power draw
38 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B
Quantization
IQ2_M
Framework
Ollama
Version
v0.53
Context length
32768 tokens
Batch size
512
GPU layers
20
Flash Attention
Off

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
Apple M1 Max
Measured VRAM (avg)
6.47 GB
OS
macOS 15.0
Driver
Apple M1 Max
CUDA
N/A
Power draw
38 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

    v0.53 · macOS 15.0 · CUDA N/A · 32768 ctx

    126.39 tok/s

    Decode

    412.92 tok/s

    Prefill

    1.07 s

    TTFT

    6.47 GB

    VRAM

    —

    MTP

    38 W

    Power draw

    L2 batch seed 2026-10-16: Ollama on hw_id=101 running model_id=54 at IQ2_M