Arao AI
Back to search results

Qwen2.5-0.5B-Instruct · IQ2_XS

Start a discussion
NVIDIA DGX Spark128GBllama.cpp v0.51L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-0.5B-Instruct (IQ2_XS) on NVIDIA DGX Spark with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

344.66 tok/s

Decode

Decode speed

1,710.46 tok/s

Prefill

Prefill speed

0.34 s

TTFT

Time to first token

2.6 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 · IQ2_XS (current)Decode 344.66 · Prefill 1710.46 ·
  2. Strata · IQ2_XS Decode 295.77 · Prefill 1437.09 ·
  3. Strata · Q5_K_M Decode 294.11 · Prefill 1222.22 ·
  4. Ollama · IQ4_XS Decode 345.19 · Prefill 1510.37 ·
  5. Ollama · Q6_K Decode 298.04 · Prefill 1558.09 ·

Core figures

Decode (avg)
344.66 tok/s
Prefill (avg)
1,710.46 tok/s
TTFT (avg)
0.34 s
VRAM (avg)
2.6 GB
MTP acceptance rate
—
TTFB
2.76 GB
Power draw
273.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-0.5B-Instruct
Quantization
IQ2_XS
Framework
llama.cpp
Version
v0.51
Context length
8192 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
NVIDIA DGX Spark
Nominal VRAM
128 GB
Measured VRAM (avg)
2.6 GB
OS
Windows 11 23H2
Driver
560.94
CUDA
12.6
Power draw
273.7 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

    v0.51 · Windows 11 23H2 · CUDA 12.6 · 8192 ctx

    344.66 tok/s

    Decode

    1,710.46 tok/s

    Prefill

    0.34 s

    TTFT

    2.6 GB

    VRAM

    —

    MTP

    273.7 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=54 running model_id=59 at IQ2_XS