Arao AI
Back to search results

Qwen3.8-9B · IQ4_XS

Start a discussion
NVIDIA RTX 508016GBllama.cpp v0.25L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.8-9B (IQ4_XS) on NVIDIA RTX 5080 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

97.14 tok/s

Decode

Decode speed

496.38 tok/s

Prefill

Prefill speed

0.95 s

TTFT

Time to first token

15.2 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · IQ4_XS (current)Decode 97.14 · Prefill 496.38 ·
  2. llama.cpp · Q3_K_M Decode 108.92 · Prefill 455.22 ·
  3. vLLM · IQ2_M Decode 131.31 · Prefill 559.09 ·

Core figures

Decode (avg)
97.14 tok/s
Prefill (avg)
496.38 tok/s
TTFT (avg)
0.95 s
VRAM (avg)
15.2 GB
MTP acceptance rate
—
TTFB
0.78 GB
Power draw
315.9 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.8-9B
Quantization
IQ4_XS
Framework
llama.cpp
Version
v0.25
Context length
8192 tokens
Batch size
256
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)
15.2 GB
OS
macOS 15.0
Driver
555.85
CUDA
12.8
Power draw
315.9 W

Sources & evidence

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

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

    v0.25 · macOS 15.0 · CUDA 12.8 · 8192 ctx

    97.14 tok/s

    Decode

    496.38 tok/s

    Prefill

    0.95 s

    TTFT

    15.2 GB

    VRAM

    —

    MTP

    315.9 W

    Power draw

    L2 batch seed 2026-10-10: llama.cpp on hw_id=14 running model_id=52 at IQ4_XS