Arao AI
Back to search results

gemma-4-12B-it · IQ3_XS

Start a discussion
NVIDIA RTX 409024GBllama.cpp v0.55L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-12B-it (IQ3_XS) on NVIDIA RTX 4090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

114.16 tok/s

Decode

Decode speed

511.31 tok/s

Prefill

Prefill speed

1.89 s

TTFT

Time to first token

17.72 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · IQ3_XS (current)Decode 114.16 · Prefill 511.31 ·
  2. llama.cpp · Q4_K_S Decode 100.24 · Prefill 360.86 ·
  3. vLLM · Q2_K Decode 139.41 · Prefill 531.29 ·
  4. Ollama · IQ4_XS Decode 95.26 · Prefill 447.14 ·
  5. Ollama · Q6_K Decode 71.51 · Prefill 296.42 ·
  6. vLLM · Q8_0 Decode 65.9 · Prefill 494.3 ·
  7. vLLM · Q5_K_M Decode 96.3 · Prefill 722.3 ·
  8. vLLM · Q4_K_M Decode 117.4 · Prefill 880.5 ·
  9. llama.cpp · Q8_0 Decode 60.5 · Prefill 453.8 ·
  10. llama.cpp · Q5_K_M Decode 88.3 · Prefill 662.3 ·
  11. llama.cpp · Q4_K_M Decode 107.7 · Prefill 807.8 ·

Core figures

Decode (avg)
114.16 tok/s
Prefill (avg)
511.31 tok/s
TTFT (avg)
1.89 s
VRAM (avg)
17.72 GB
MTP acceptance rate
—
TTFB
0.91 GB
Power draw
376.8 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-12B-it
Quantization
IQ3_XS
Framework
llama.cpp
Version
v0.55
Context length
16384 tokens
Batch size
256
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 4090
Nominal VRAM
24 GB
Measured VRAM (avg)
17.72 GB
OS
Windows 11 23H2
Driver
570.86
CUDA
12.4
Power draw
376.8 W

Sources & evidence

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

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

    v0.55 · Windows 11 23H2 · CUDA 12.4 · 16384 ctx

    114.16 tok/s

    Decode

    511.31 tok/s

    Prefill

    1.89 s

    TTFT

    17.72 GB

    VRAM

    —

    MTP

    376.8 W

    Power draw

    L2 batch seed 2026-10-10: llama.cpp on hw_id=1 running model_id=65 at IQ3_XS