Arao AI
Back to search results

Qwen3-8B · IQ2_M

Start a discussion
NVIDIA RTX 3060 12GB12GBllama.cpp v0.55L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-8B (IQ2_M) on NVIDIA RTX 3060 12GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

70.85 tok/s

Decode

Decode speed

298.67 tok/s

Prefill

Prefill speed

0.13 s

TTFT

Time to first token

10.84 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · IQ2_M (current)Decode 70.85 · Prefill 298.67 ·
  2. vLLM · IQ3_S Decode 64 · Prefill 232.21 ·
  3. llama.cpp · IQ3_S Decode 81.43 · Prefill 366.11 ·

Core figures

Decode (avg)
70.85 tok/s
Prefill (avg)
298.67 tok/s
TTFT (avg)
0.13 s
VRAM (avg)
10.84 GB
MTP acceptance rate
—
TTFB
0.57 GB
Power draw
163.6 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-8B
Quantization
IQ2_M
Framework
llama.cpp
Version
v0.55
Context length
32768 tokens
Batch size
512
GPU layers
33
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)
10.84 GB
OS
Ubuntu 22.04
Driver
560.94
CUDA
12.6
Power draw
163.6 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-18

    v0.55 · Ubuntu 22.04 · CUDA 12.6 · 32768 ctx

    70.85 tok/s

    Decode

    298.67 tok/s

    Prefill

    0.13 s

    TTFT

    10.84 GB

    VRAM

    —

    MTP

    163.6 W

    Power draw

    L2 batch seed 2026-10-18: llama.cpp on hw_id=15 running model_id=53 at IQ2_M