Arao AI
Back to search results

Qwen3.5-0.8B · IQ4_XS

Start a discussion
NVIDIA RTX 507012GBvLLM v0.36L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-0.8B (IQ4_XS) on NVIDIA RTX 5070 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

188.25 tok/s

Decode

Decode speed

873.82 tok/s

Prefill

Prefill speed

0.97 s

TTFT

Time to first token

3.34 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. vLLM · IQ3_XS Decode 223.3 · Prefill 1047.53 ·
  2. vLLM · IQ4_XS (current)Decode 188.25 · Prefill 873.82 ·
  3. Ollama · Q3_K_S Decode 197.53 · Prefill 685.98 ·
  4. Ollama · Q5_K_S Decode 124.72 · Prefill 677.51 ·
  5. Strata · IQ4_XS Decode 233.02 · Prefill 915.89 ·
  6. Strata · Q6_K Decode 189.01 · Prefill 746.35 ·
  7. llama.cpp · Q8_0 Decode 582.4 · Prefill 4368 ·

Core figures

Decode (avg)
188.25 tok/s
Prefill (avg)
873.82 tok/s
TTFT (avg)
0.97 s
VRAM (avg)
3.34 GB
MTP acceptance rate
—
TTFB
1.51 GB
Power draw
184 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-0.8B
Quantization
IQ4_XS
Framework
vLLM
Version
v0.36
Context length
2048 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 RTX 5070
Nominal VRAM
12 GB
Measured VRAM (avg)
3.34 GB
OS
Ubuntu 22.04
Driver
560.94
CUDA
12.8
Power draw
184 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

    v0.36 · Ubuntu 22.04 · CUDA 12.8 · 2048 ctx

    188.25 tok/s

    Decode

    873.82 tok/s

    Prefill

    0.97 s

    TTFT

    3.34 GB

    VRAM

    —

    MTP

    184 W

    Power draw

    L2 batch seed 2026-10-10: vLLM on hw_id=18 running model_id=51 at IQ4_XS