Arao AI
Back to search results

Qwen3-4B · Q4_K_M

Start a discussion
NVIDIA RTX 4070 Ti12GBllama.cpp v0.53L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B (Q4_K_M) on NVIDIA RTX 4070 Ti with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

104 tok/s

Decode

Decode speed

443.35 tok/s

Prefill

Prefill speed

1.2 s

TTFT

Time to first token

8.99 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 · Q3_K_L Decode 107.42 · Prefill 453.38 ·
  2. llama.cpp · Q4_K_M (current)Decode 104 · Prefill 443.35 ·
  3. vLLM · Q8_0 Decode 66.34 · Prefill 241.62 ·

Core figures

Decode (avg)
104 tok/s
Prefill (avg)
443.35 tok/s
TTFT (avg)
1.2 s
VRAM (avg)
8.99 GB
MTP acceptance rate
—
TTFB
0.83 GB
Power draw
184.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B
Quantization
Q4_K_M
Framework
llama.cpp
Version
v0.53
Context length
8192 tokens
Batch size
512
GPU layers
33
Flash Attention
Off

Hardware

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

GPU
NVIDIA RTX 4070 Ti
Nominal VRAM
12 GB
Measured VRAM (avg)
8.99 GB
OS
Ubuntu 24.04
Driver
570.86
CUDA
12.6
Power draw
184.5 W

Sources & evidence

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

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

    v0.53 · Ubuntu 24.04 · CUDA 12.6 · 8192 ctx

    104 tok/s

    Decode

    443.35 tok/s

    Prefill

    1.2 s

    TTFT

    8.99 GB

    VRAM

    —

    MTP

    184.5 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=42 running model_id=54 at Q4_K_M