Arao AI
Back to search results

Qwen3.8-9B · IQ1_S

Start a discussion
NVIDIA RTX 507012GBvLLM v0.20L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.8-9B (IQ1_S) on NVIDIA RTX 5070 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

119.3 tok/s

Decode

Decode speed

428.02 tok/s

Prefill

Prefill speed

0.7 s

TTFT

Time to first token

10.04 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. vLLM · IQ1_S (current)Decode 119.3 · Prefill 428.02 ·
  2. Strata · IQ1_S Decode 153.3 · Prefill 756.44 ·
  3. Strata · IQ2_XS Decode 135.12 · Prefill 577.01 ·
  4. Ollama · IQ1_S Decode 93.91 · Prefill 396.52 ·
  5. Strata · IQ4_XS Decode 137.1 · Prefill 1028.3 ·

Core figures

Decode (avg)
119.3 tok/s
Prefill (avg)
428.02 tok/s
TTFT (avg)
0.7 s
VRAM (avg)
10.04 GB
MTP acceptance rate
—
TTFB
0.95 GB
Power draw
226.1 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.8-9B
Quantization
IQ1_S
Framework
vLLM
Version
v0.20
Context length
8192 tokens
Batch size
16
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 5070
Nominal VRAM
12 GB
Measured VRAM (avg)
10.04 GB
OS
Ubuntu 22.04
Driver
570.86
CUDA
12.4
Power draw
226.1 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

    v0.20 · Ubuntu 22.04 · CUDA 12.4 · 8192 ctx

    119.3 tok/s

    Decode

    428.02 tok/s

    Prefill

    0.7 s

    TTFT

    10.04 GB

    VRAM

    —

    MTP

    226.1 W

    Power draw

    L2 batch seed 2026-10-18: vLLM on hw_id=18 running model_id=52 at IQ1_S