Arao AI
Back to search results

Qwen3-8B · Q4_K_S

Start a discussion
NVIDIA Jetson AGX Orin 64GB64GBvLLM v0.49L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-8B (Q4_K_S) on NVIDIA Jetson AGX Orin 64GB with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

62 tok/s

Decode

Decode speed

313.33 tok/s

Prefill

Prefill speed

0.22 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

5 days ago

Last verified

Performance

  1. llama.cpp · IQ1_S Decode 80.69 · Prefill 386.89 ·
  2. llama.cpp · IQ2_XS Decode 81.76 · Prefill 342.2 ·
  3. Ollama · FP16 Decode 24.64 · Prefill 90.65 ·
  4. Ollama · Q3_K_L Decode 56.02 · Prefill 230.05 ·
  5. vLLM · Q3_K_S Decode 73.73 · Prefill 262.59 ·
  6. vLLM · Q4_K_S (current)Decode 62 · Prefill 313.33 ·

Core figures

Decode (avg)
62 tok/s
Prefill (avg)
313.33 tok/s
TTFT (avg)
0.22 s
VRAM (avg)
15.2 GB
MTP acceptance rate
—
TTFB
0.5 GB
Power draw
25.9 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-8B
Quantization
Q4_K_S
Framework
vLLM
Version
v0.49
Context length
8192 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 Jetson AGX Orin 64GB
Nominal VRAM
64 GB
Measured VRAM (avg)
15.2 GB
OS
JetPack 6.0
Driver
NVIDIA L4T 36.3
CUDA
12.6
Power draw
25.9 W

Sources & evidence

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

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

    v0.49 · JetPack 6.0 · CUDA 12.6 · 8192 ctx

    62 tok/s

    Decode

    313.33 tok/s

    Prefill

    0.22 s

    TTFT

    15.2 GB

    VRAM

    —

    MTP

    25.9 W

    Power draw

    L2 batch seed 2026-10-13: vLLM on hw_id=76 running model_id=53 at Q4_K_S