Arao AI
Back to search results

Qwen3.5-0.8B · IQ2_M

Start a discussion
NVIDIA Jetson AGX Orin 64GB64GBllama.cpp v0.38L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-0.8B (IQ2_M) on NVIDIA Jetson AGX Orin 64GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

177.54 tok/s

Decode

Decode speed

614.93 tok/s

Prefill

Prefill speed

1.84 s

TTFT

Time to first token

2.97 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. llama.cpp · IQ2_M (current)Decode 177.54 · Prefill 614.93 ·
  2. llama.cpp · IQ2_XS Decode 200.96 · Prefill 773.99 ·
  3. Ollama · IQ3_S Decode 138.5 · Prefill 456.82 ·
  4. Ollama · Q5_K_S Decode 116.4 · Prefill 439.62 ·
  5. vLLM · IQ4_XS Decode 140.75 · Prefill 759.45 ·
  6. vLLM · Q5_K_M Decode 125.95 · Prefill 424.45 ·

Core figures

Decode (avg)
177.54 tok/s
Prefill (avg)
614.93 tok/s
TTFT (avg)
1.84 s
VRAM (avg)
2.97 GB
MTP acceptance rate
—
TTFB
1.42 GB
Power draw
25.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-0.8B
Quantization
IQ2_M
Framework
llama.cpp
Version
v0.38
Context length
8192 tokens
Batch size
512
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 Jetson AGX Orin 64GB
Nominal VRAM
64 GB
Measured VRAM (avg)
2.97 GB
OS
JetPack 6.0
Driver
NVIDIA L4T 36.3
CUDA
12.6
Power draw
25.5 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-16

    v0.38 · JetPack 6.0 · CUDA 12.6 · 8192 ctx

    177.54 tok/s

    Decode

    614.93 tok/s

    Prefill

    1.84 s

    TTFT

    2.97 GB

    VRAM

    —

    MTP

    25.5 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=76 running model_id=51 at IQ2_M