Arao AI
Back to search results

gemma-4-31B-it · IQ2_M

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.53L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-31B-it (IQ2_M) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

24.41 tok/s

Decode

Decode speed

110.5 tok/s

Prefill

Prefill speed

0.82 s

TTFT

Time to first token

35.95 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

17 days ago

Last verified

Performance

  1. llama.cpp · IQ2_M (current)Decode 24.41 · Prefill 110.5 ·
  2. llama.cpp · IQ2_XS Decode 25.14 · Prefill 89.05 ·
  3. Ollama · IQ2_M Decode 24.86 · Prefill 123.01 ·
  4. Strata · IQ3_XS Decode 81.29 · Prefill 414.62 ·
  5. Strata · Q3_K_S Decode 82.27 · Prefill 338.79 ·

Core figures

Decode (avg)
24.41 tok/s
Prefill (avg)
110.5 tok/s
TTFT (avg)
0.82 s
VRAM (avg)
35.95 GB
MTP acceptance rate
—
TTFB
0.2 GB
Power draw
60.3 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-31B-it
Quantization
IQ2_M
Framework
llama.cpp
Version
v0.53
Context length
32768 tokens
Batch size
512
GPU layers
99
Flash Attention
Off

Hardware

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

GPU
AMD Ryzen AI Max+ 395
Nominal VRAM
128 GB
Measured VRAM (avg)
35.95 GB
OS
Ubuntu 24.04
Driver
amdgpu 6.11
CUDA
12.6
Power draw
60.3 W

Sources & evidence

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

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

    v0.53 · Ubuntu 24.04 · CUDA 12.6 · 32768 ctx

    24.41 tok/s

    Decode

    110.5 tok/s

    Prefill

    0.82 s

    TTFT

    35.95 GB

    VRAM

    —

    MTP

    60.3 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=24 running model_id=63 at IQ2_M