Arao AI
Back to search results

Qwen3.5-4B · IQ1_S

Start a discussion
NVIDIA RTX 508016GBllama.cpp v0.45L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-4B (IQ1_S) on NVIDIA RTX 5080 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

350.61 tok/s

Decode

Decode speed

1,829.73 tok/s

Prefill

Prefill speed

1.22 s

TTFT

Time to first token

5.63 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · IQ1_S (current)Decode 350.61 · Prefill 1829.73 ·
  2. llama.cpp · IQ2_XS Decode 312.6 · Prefill 1217.96 ·
  3. Strata · IQ3_S Decode 356.3 · Prefill 1901.07 ·
  4. llama.cpp · IQ4_XS Decode 133.14 · Prefill 731.5 ·
  5. Strata · Q3_K_M Decode 231.73 · Prefill 785.76 ·

Core figures

Decode (avg)
350.61 tok/s
Prefill (avg)
1,829.73 tok/s
TTFT (avg)
1.22 s
VRAM (avg)
5.63 GB
MTP acceptance rate
—
TTFB
2.8 GB
Power draw
314.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-4B
Quantization
IQ1_S
Framework
llama.cpp
Version
v0.45
Context length
32768 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)
5.63 GB
OS
Windows 11 24H2
Driver
570.86
CUDA
13.0
Power draw
314.7 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

    v0.45 · Windows 11 24H2 · CUDA 13.0 · 32768 ctx

    350.61 tok/s

    Decode

    1,829.73 tok/s

    Prefill

    1.22 s

    TTFT

    5.63 GB

    VRAM

    —

    MTP

    314.7 W

    Power draw

    L2 batch seed 2026-10-18: llama.cpp on hw_id=14 running model_id=48 at IQ1_S