Arao AI
Back to search results

Qwen3-4B · Q4_K_M

Start a discussion
NVIDIA RTX 407012GBvLLM v0.29L2 Cross-framework verified

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

88.36 tok/s

Decode

Decode speed

459.98 tok/s

Prefill

Prefill speed

1.31 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. vLLM · Q3_K_S Decode 110.14 · Prefill 421.19 ·
  2. vLLM · Q4_K_M (current)Decode 88.36 · Prefill 459.98 ·

Core figures

Decode (avg)
88.36 tok/s
Prefill (avg)
459.98 tok/s
TTFT (avg)
1.31 s
VRAM (avg)
8.99 GB
MTP acceptance rate
—
TTFB
0.71 GB
Power draw
258.8 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B
Quantization
Q4_K_M
Framework
vLLM
Version
v0.29
Context length
32768 tokens
Batch size
32
GPU layers
999
Flash Attention
On

Hardware

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

GPU
NVIDIA RTX 4070
Nominal VRAM
12 GB
Measured VRAM (avg)
8.99 GB
OS
macOS 15.0
Driver
560.94
CUDA
12.8
Power draw
258.8 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

    v0.29 · macOS 15.0 · CUDA 12.8 · 32768 ctx

    88.36 tok/s

    Decode

    459.98 tok/s

    Prefill

    1.31 s

    TTFT

    8.99 GB

    VRAM

    —

    MTP

    258.8 W

    Power draw

    L2 batch seed 2026-10-13: vLLM on hw_id=45 running model_id=54 at Q4_K_M