Arao AI
Back to search results

Ministral-3-8B-Instruct-2512 · Q3_K_L

Start a discussion
AMD Ryzen AI Max+ 395128GBOllama v0.37L2 Cross-framework verified

This page aggregates 1 real-world runs of Ministral-3-8B-Instruct-2512 (Q3_K_L) on AMD Ryzen AI Max+ 395 with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.

32.47 tok/s

Decode

Decode speed

141.58 tok/s

Prefill

Prefill speed

1.39 s

TTFT

Time to first token

14.2 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

11 days ago

Last verified

Performance

  1. llama.cpp · Q2_K Decode 42.14 · Prefill 173.71 ·
  2. Ollama · Q3_K_L (current)Decode 32.47 · Prefill 141.58 ·

Core figures

Decode (avg)
32.47 tok/s
Prefill (avg)
141.58 tok/s
TTFT (avg)
1.39 s
VRAM (avg)
14.2 GB
MTP acceptance rate
—
TTFB
0.26 GB
Power draw
49.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Ministral-3-8B-Instruct-2512
Quantization
Q3_K_L
Framework
Ollama
Version
v0.37
Context length
16384 tokens
Batch size
256
GPU layers
20
Flash Attention
On

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)
14.2 GB
OS
Windows 11 23H2
Driver
amdgpu 6.12
CUDA
12.4
Power draw
49.3 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-24

    v0.37 · Windows 11 23H2 · CUDA 12.4 · 16384 ctx

    32.47 tok/s

    Decode

    141.58 tok/s

    Prefill

    1.39 s

    TTFT

    14.2 GB

    VRAM

    —

    MTP

    49.3 W

    Power draw

    L2 batch seed 2026-10-24: Ollama on hw_id=24 running model_id=74 at Q3_K_L