Arao AI
Back to search results

Llama-3.2-3B-Instruct · Q5_K_S

Start a discussion
NVIDIA RTX 508016GBllama.cpp v0.42L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (Q5_K_S) on NVIDIA RTX 5080 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

381.8 tok/s

Decode

Decode speed

1,978.14 tok/s

Prefill

Prefill speed

1.28 s

TTFT

Time to first token

7.77 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 · Q5_K_S (current)Decode 381.8 · Prefill 1978.14 ·
  2. Ollama · Q5_K_S Decode 312.17 · Prefill 1673.45 ·
  3. Ollama · Q8_0 Decode 294.03 · Prefill 1167.26 ·
  4. llama.cpp · IQ1_S Decode 351.28 · Prefill 1853.09 ·
  5. Ollama · IQ4_XS Decode 362.6 · Prefill 1325.05 ·

Core figures

Decode (avg)
381.8 tok/s
Prefill (avg)
1,978.14 tok/s
TTFT (avg)
1.28 s
VRAM (avg)
7.77 GB
MTP acceptance rate
—
TTFB
3.05 GB
Power draw
281.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.2-3B-Instruct
Quantization
Q5_K_S
Framework
llama.cpp
Version
v0.42
Context length
2048 tokens
Batch size
512
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
NVIDIA RTX 5080
Nominal VRAM
16 GB
Measured VRAM (avg)
7.77 GB
OS
Windows 11 24H2
Driver
570.86
CUDA
12.6
Power draw
281.4 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

    v0.42 · Windows 11 24H2 · CUDA 12.6 · 2048 ctx

    381.8 tok/s

    Decode

    1,978.14 tok/s

    Prefill

    1.28 s

    TTFT

    7.77 GB

    VRAM

    —

    MTP

    281.4 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=14 running model_id=70 at Q5_K_S