Arao AI
Back to search results

Qwen3.5-0.8B · FP16

Start a discussion
NVIDIA RTX 3060 12GB12GBllama.cpp v0.26L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-0.8B (FP16) on NVIDIA RTX 3060 12GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

72.92 tok/s

Decode

Decode speed

274.88 tok/s

Prefill

Prefill speed

1.59 s

TTFT

Time to first token

5.21 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 · FP16 (current)Decode 72.92 · Prefill 274.88 ·
  2. llama.cpp · Q5_K_S Decode 132.45 · Prefill 600.24 ·
  3. Ollama · Q3_K_S Decode 138.92 · Prefill 607.45 ·
  4. Ollama · FP16 Decode 65.06 · Prefill 244.73 ·
  5. Ollama · Q2_K Decode 166.72 · Prefill 642.75 ·
  6. Strata · FP16 Decode 103.42 · Prefill 466.54 ·
  7. llama.cpp · Q3_K_S Decode 187.82 · Prefill 792.93 ·
  8. Ollama · Q3_K_L Decode 167.74 · Prefill 790.34 ·
  9. Ollama · Q5_K_M Decode 118.62 · Prefill 422.11 ·
  10. Strata · Q3_K_M Decode 279.25 · Prefill 1294.74 ·
  11. Strata · Q5_K_S Decode 214.78 · Prefill 1075.84 ·
  12. llama.cpp · Q8_0 Decode 312 · Prefill 2340 ·

Core figures

Decode (avg)
72.92 tok/s
Prefill (avg)
274.88 tok/s
TTFT (avg)
1.59 s
VRAM (avg)
5.21 GB
MTP acceptance rate
—
TTFB
0.58 GB
Power draw
144.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-0.8B
Quantization
FP16
Framework
llama.cpp
Version
v0.26
Context length
4096 tokens
Batch size
256
GPU layers
99
Flash Attention
On

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
NVIDIA RTX 3060 12GB
Nominal VRAM
12 GB
Measured VRAM (avg)
5.21 GB
OS
Windows 11 24H2
Driver
565.57
CUDA
12.8
Power draw
144.7 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

    v0.26 · Windows 11 24H2 · CUDA 12.8 · 4096 ctx

    72.92 tok/s

    Decode

    274.88 tok/s

    Prefill

    1.59 s

    TTFT

    5.21 GB

    VRAM

    —

    MTP

    144.7 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=15 running model_id=51 at FP16