Arao AI
Back to search results

Qwen3.5-4B · Q3_K_L

Start a discussion
NVIDIA RTX 509032GBllama.cpp v0.30L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-4B (Q3_K_L) on NVIDIA RTX 5090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

338.21 tok/s

Decode

Decode speed

1,409.78 tok/s

Prefill

Prefill speed

2.14 s

TTFT

Time to first token

8.15 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · Q3_K_L (current)Decode 338.21 · Prefill 1409.78 ·
  2. vLLM · Q3_K_M Decode 356.1 · Prefill 1633.97 ·
  3. vLLM · Q3_K_S Decode 324.89 · Prefill 1253.12 ·
  4. Strata · Q4_K_S Decode 353.67 · Prefill 1676.75 ·

Core figures

Decode (avg)
338.21 tok/s
Prefill (avg)
1,409.78 tok/s
TTFT (avg)
2.14 s
VRAM (avg)
8.15 GB
MTP acceptance rate
—
TTFB
2.71 GB
Power draw
431.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-4B
Quantization
Q3_K_L
Framework
llama.cpp
Version
v0.30
Context length
8192 tokens
Batch size
512
GPU layers
99
Flash Attention
Off

Hardware

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

GPU
NVIDIA RTX 5090
Nominal VRAM
32 GB
Measured VRAM (avg)
8.15 GB
OS
Ubuntu 24.04
Driver
555.85
CUDA
12.8
Power draw
431.5 W

Sources & evidence

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

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

    v0.30 · Ubuntu 24.04 · CUDA 12.8 · 8192 ctx

    338.21 tok/s

    Decode

    1,409.78 tok/s

    Prefill

    2.14 s

    TTFT

    8.15 GB

    VRAM

    —

    MTP

    431.5 W

    Power draw

    L2 batch seed 2026-10-10: llama.cpp on hw_id=13 running model_id=48 at Q3_K_L