Arao AI
Back to search results

Qwen3-1.7B · IQ2_M

Start a discussion
NVIDIA RTX 409024GBllama.cpp v0.33L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-1.7B (IQ2_M) on NVIDIA RTX 4090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

323.97 tok/s

Decode

Decode speed

1,188.09 tok/s

Prefill

Prefill speed

2.05 s

TTFT

Time to first token

3.96 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 323.97 · Prefill 1188.09 ·
  2. llama.cpp · Q2_K Decode 306.65 · Prefill 1436.24 ·
  3. Strata · IQ1_S Decode 287.87 · Prefill 1066.63 ·
  4. Strata · Q2_K Decode 285.75 · Prefill 1116.76 ·

Core figures

Decode (avg)
323.97 tok/s
Prefill (avg)
1,188.09 tok/s
TTFT (avg)
2.05 s
VRAM (avg)
3.96 GB
MTP acceptance rate
—
TTFB
2.59 GB
Power draw
470.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-1.7B
Quantization
IQ2_M
Framework
llama.cpp
Version
v0.33
Context length
32768 tokens
Batch size
512
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 4090
Nominal VRAM
24 GB
Measured VRAM (avg)
3.96 GB
OS
Windows 11 23H2
Driver
570.86
CUDA
12.6
Power draw
470.3 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

    v0.33 · Windows 11 23H2 · CUDA 12.6 · 32768 ctx

    323.97 tok/s

    Decode

    1,188.09 tok/s

    Prefill

    2.05 s

    TTFT

    3.96 GB

    VRAM

    —

    MTP

    470.3 W

    Power draw

    L2 batch seed 2026-10-10: llama.cpp on hw_id=1 running model_id=55 at IQ2_M