Arao AI
Back to search results

Qwen2.5-0.5B-Instruct · IQ3_XS

Start a discussion
NVIDIA RTX 509032GBvLLM v0.25L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-0.5B-Instruct (IQ3_XS) on NVIDIA RTX 5090 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

285.37 tok/s

Decode

Decode speed

1,069.59 tok/s

Prefill

Prefill speed

1.76 s

TTFT

Time to first token

2.75 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. llama.cpp · IQ4_XS Decode 319.07 · Prefill 1265.71 ·
  2. llama.cpp · Q8_0 Decode 358.97 · Prefill 1824.25 ·
  3. vLLM · IQ3_XS (current)Decode 285.37 · Prefill 1069.59 ·

Core figures

Decode (avg)
285.37 tok/s
Prefill (avg)
1,069.59 tok/s
TTFT (avg)
1.76 s
VRAM (avg)
2.75 GB
MTP acceptance rate
—
TTFB
2.28 GB
Power draw
477.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-0.5B-Instruct
Quantization
IQ3_XS
Framework
vLLM
Version
v0.25
Context length
2048 tokens
Batch size
8
GPU layers
999
Flash Attention
On

Hardware

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

GPU
NVIDIA RTX 5090
Nominal VRAM
32 GB
Measured VRAM (avg)
2.75 GB
OS
Windows 11 23H2
Driver
560.94
CUDA
12.8
Power draw
477.7 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-16

    v0.25 · Windows 11 23H2 · CUDA 12.8 · 2048 ctx

    285.37 tok/s

    Decode

    1,069.59 tok/s

    Prefill

    1.76 s

    TTFT

    2.75 GB

    VRAM

    —

    MTP

    477.7 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=13 running model_id=59 at IQ3_XS