Arao AI
Back to search results

Flint-40B-A2.5B · IQ3_XS

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.30L2 Cross-framework verified

This page aggregates 1 real-world runs of Flint-40B-A2.5B (IQ3_XS) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

77.09 tok/s

Decode

Decode speed

382.38 tok/s

Prefill

Prefill speed

0.6 s

TTFT

Time to first token

53.92 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

14 days ago

Last verified

Performance

  1. llama.cpp · IQ3_XS (current)Decode 77.09 · Prefill 382.38 ·
  2. vLLM · Q4_K_S Decode 63.8 · Prefill 213.74 ·

Core figures

Decode (avg)
77.09 tok/s
Prefill (avg)
382.38 tok/s
TTFT (avg)
0.6 s
VRAM (avg)
53.92 GB
MTP acceptance rate
—
TTFB
0.62 GB
Power draw
46.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Flint-40B-A2.5B
Quantization
IQ3_XS
Framework
llama.cpp
Version
v0.30
Context length
2048 tokens
Batch size
512
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
AMD Ryzen AI Max+ 395
Nominal VRAM
128 GB
Measured VRAM (avg)
53.92 GB
OS
Ubuntu 22.04
Driver
amdgpu 6.11
CUDA
12.4
Power draw
46.7 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 2026-10-22

    v0.30 · Ubuntu 22.04 · CUDA 12.4 · 2048 ctx

    77.09 tok/s

    Decode

    382.38 tok/s

    Prefill

    0.6 s

    TTFT

    53.92 GB

    VRAM

    —

    MTP

    46.7 W

    Power draw

    L2 batch seed 2026-10-22: llama.cpp on hw_id=24 running model_id=92 at IQ3_XS