Arao AI
Back to search results

Qwen3.5-9B · Q4_K_M

Start a discussion
NVIDIA Jetson AGX Orin 64GB64GBllama.cpp v0.52L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-9B (Q4_K_M) on NVIDIA Jetson AGX Orin 64GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

47.21 tok/s

Decode

Decode speed

254.51 tok/s

Prefill

Prefill speed

0.65 s

TTFT

Time to first token

17.6 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. Ollama · Q3_K_S Decode 55.39 · Prefill 272.09 ·
  2. Strata · Q4_K_M Decode 72.01 · Prefill 281.13 ·
  3. llama.cpp · IQ2_XS Decode 71.62 · Prefill 230.5 ·
  4. llama.cpp · Q4_K_M (current)Decode 47.21 · Prefill 254.51 ·
  5. Strata · Q4_K_S Decode 76.44 · Prefill 335.52 ·
  6. Strata · Q6_K Decode 51.24 · Prefill 260.03 ·

Core figures

Decode (avg)
47.21 tok/s
Prefill (avg)
254.51 tok/s
TTFT (avg)
0.65 s
VRAM (avg)
17.6 GB
MTP acceptance rate
—
TTFB
0.38 GB
Power draw
58.2 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-9B
Quantization
Q4_K_M
Framework
llama.cpp
Version
v0.52
Context length
8192 tokens
Batch size
512
GPU layers
99
Flash Attention
Off

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)
17.6 GB
OS
JetPack 6.0
Driver
NVIDIA L4T 36.3
CUDA
12.6
Power draw
58.2 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

    v0.52 · JetPack 6.0 · CUDA 12.6 · 8192 ctx

    47.21 tok/s

    Decode

    254.51 tok/s

    Prefill

    0.65 s

    TTFT

    17.6 GB

    VRAM

    —

    MTP

    58.2 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=76 running model_id=49 at Q4_K_M