Arao AI
Back to search results

Llama-3.2-3B-Instruct · IQ3_XS

Start a discussion
AMD Radeon RX 9070 XT16GBllama.cpp v0.48L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (IQ3_XS) on AMD Radeon RX 9070 XT with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

319.75 tok/s

Decode

Decode speed

1,663.47 tok/s

Prefill

Prefill speed

1.88 s

TTFT

Time to first token

6.01 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · IQ3_XS (current)Decode 319.75 · Prefill 1663.47 ·
  2. Strata · IQ3_S Decode 308.2 · Prefill 1415.95 ·
  3. Strata · IQ4_XS Decode 311.94 · Prefill 1321.43 ·

Core figures

Decode (avg)
319.75 tok/s
Prefill (avg)
1,663.47 tok/s
TTFT (avg)
1.88 s
VRAM (avg)
6.01 GB
MTP acceptance rate
—
TTFB
2.56 GB
Power draw
200.6 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.2-3B-Instruct
Quantization
IQ3_XS
Framework
llama.cpp
Version
v0.48
Context length
4096 tokens
Batch size
256
GPU layers
33
Flash Attention
On

Hardware

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

GPU
AMD Radeon RX 9070 XT
Nominal VRAM
16 GB
Measured VRAM (avg)
6.01 GB
OS
Ubuntu 24.04
Driver
amdgpu 6.11
CUDA
12.6
Power draw
200.6 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-18

    v0.48 · Ubuntu 24.04 · CUDA 12.6 · 4096 ctx

    319.75 tok/s

    Decode

    1,663.47 tok/s

    Prefill

    1.88 s

    TTFT

    6.01 GB

    VRAM

    —

    MTP

    200.6 W

    Power draw

    L2 batch seed 2026-10-18: llama.cpp on hw_id=16 running model_id=70 at IQ3_XS