Arao AI
Back to search results

Qwen2.5-0.5B-Instruct · IQ1_S

Start a discussion
NVIDIA RTX 407012GBllama.cpp v0.44L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-0.5B-Instruct (IQ1_S) on NVIDIA RTX 4070 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

330.62 tok/s

Decode

Decode speed

1,805.64 tok/s

Prefill

Prefill speed

2.46 s

TTFT

Time to first token

2.54 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. llama.cpp · IQ1_S (current)Decode 330.62 · Prefill 1805.64 ·
  2. llama.cpp · Q5_K_S Decode 179.43 · Prefill 799.81 ·
  3. vLLM · IQ2_M Decode 318.42 · Prefill 1565.27 ·
  4. vLLM · Q3_K_M Decode 277.32 · Prefill 1224.46 ·

Core figures

Decode (avg)
330.62 tok/s
Prefill (avg)
1,805.64 tok/s
TTFT (avg)
2.46 s
VRAM (avg)
2.54 GB
MTP acceptance rate
—
TTFB
2.64 GB
Power draw
198.8 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-0.5B-Instruct
Quantization
IQ1_S
Framework
llama.cpp
Version
v0.44
Context length
32768 tokens
Batch size
256
GPU layers
99
Flash Attention
On

Hardware

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

GPU
NVIDIA RTX 4070
Nominal VRAM
12 GB
Measured VRAM (avg)
2.54 GB
OS
Ubuntu 24.04
Driver
565.57
CUDA
12.4
Power draw
198.8 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-13

    v0.44 · Ubuntu 24.04 · CUDA 12.4 · 32768 ctx

    330.62 tok/s

    Decode

    1,805.64 tok/s

    Prefill

    2.46 s

    TTFT

    2.54 GB

    VRAM

    —

    MTP

    198.8 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=45 running model_id=59 at IQ1_S