Arao AI
Back to search results

Qwen3.5-0.8B · Q8_0

Start a discussion
NVIDIA H2097GBvLLM v0.54L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-0.8B (Q8_0) on NVIDIA H20 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

82.63 tok/s

Decode

Decode speed

440.06 tok/s

Prefill

Prefill speed

0.49 s

TTFT

Time to first token

4.03 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

17 days ago

Last verified

Performance

  1. llama.cpp · Q3_K_L Decode 129.02 · Prefill 590.22 ·
  2. vLLM · Q4_K_M Decode 121.55 · Prefill 512.9 ·
  3. vLLM · Q8_0 (current)Decode 82.63 · Prefill 440.06 ·

Core figures

Decode (avg)
82.63 tok/s
Prefill (avg)
440.06 tok/s
TTFT (avg)
0.49 s
VRAM (avg)
4.03 GB
MTP acceptance rate
—
TTFB
0.66 GB
Power draw
370 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-0.8B
Quantization
Q8_0
Framework
vLLM
Version
v0.54
Context length
16384 tokens
Batch size
32
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 H20
Nominal VRAM
97 GB
Measured VRAM (avg)
4.03 GB
OS
Windows 11 23H2
Driver
565.57
CUDA
12.4
Power draw
370 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.54 · Windows 11 23H2 · CUDA 12.4 · 16384 ctx

    82.63 tok/s

    Decode

    440.06 tok/s

    Prefill

    0.49 s

    TTFT

    4.03 GB

    VRAM

    —

    MTP

    370 W

    Power draw

    L2 batch seed 2026-10-19: vLLM on hw_id=56 running model_id=51 at Q8_0