Arao AI
Back to search results

Hy-Memory-30B-A3B · IQ2_M

Start a discussion
Apple M5 Ultra512GBllama.cpp v0.36L2 Cross-framework verified

This page aggregates 1 real-world runs of Hy-Memory-30B-A3B (IQ2_M) on Apple M5 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

216.9 tok/s

Decode

Decode speed

866.69 tok/s

Prefill

Prefill speed

1.81 s

TTFT

Time to first token

34.86 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

16 days ago

Last verified

Performance

  1. llama.cpp · IQ2_M (current)Decode 216.9 · Prefill 866.69 ·
  2. llama.cpp · Q3_K_S Decode 189.37 · Prefill 679.35 ·
  3. Ollama · Q5_K_S Decode 139.67 · Prefill 562.73 ·

Core figures

Decode (avg)
216.9 tok/s
Prefill (avg)
866.69 tok/s
TTFT (avg)
1.81 s
VRAM (avg)
34.86 GB
MTP acceptance rate
—
TTFB
1.74 GB
Power draw
103.8 W

Configuration

Member-level fields are taken from the most recent run

Model
Hy-Memory-30B-A3B
Quantization
IQ2_M
Framework
llama.cpp
Version
v0.36
Context length
16384 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 Ultra
Nominal VRAM
512 GB
Measured VRAM (avg)
34.86 GB
OS
macOS 26.1
Power draw
103.8 W

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

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

    v0.36 · macOS 26.1 · 16384 ctx

    216.9 tok/s

    Decode

    866.69 tok/s

    Prefill

    1.81 s

    TTFT

    34.86 GB

    VRAM

    —

    MTP

    103.8 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=29 running model_id=87 at IQ2_M