Arao AI
Back to search results

elephant-1 · IQ3_XS

Start a discussion
NVIDIA RTX 409024GBllama.cpp v0.29L2 Cross-framework verified

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

136.35 tok/s

Decode

Decode speed

504.8 tok/s

Prefill

Prefill speed

0.59 s

TTFT

Time to first token

13.82 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

13 days ago

Last verified

Performance

  1. llama.cpp · IQ1_S Decode 180.81 · Prefill 654.72 ·
  2. llama.cpp · IQ3_XS (current)Decode 136.35 · Prefill 504.8 ·
  3. vLLM · IQ2_XS Decode 179.71 · Prefill 772.31 ·

Core figures

Decode (avg)
136.35 tok/s
Prefill (avg)
504.8 tok/s
TTFT (avg)
0.59 s
VRAM (avg)
13.82 GB
MTP acceptance rate
—
TTFB
1.09 GB
Power draw
391.8 W

Configuration

Member-level fields are taken from the most recent run

Model
elephant-1
Quantization
IQ3_XS
Framework
llama.cpp
Version
v0.29
Context length
2048 tokens
Batch size
256
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)
13.82 GB
OS
Windows 11 23H2
Driver
570.86
CUDA
12.4
Power draw
391.8 W

Get started

Original model resources

Base model resources may not match this quantization. Use the verified deployment weights above when available.

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-22

    v0.29 · Windows 11 23H2 · CUDA 12.4 · 2048 ctx

    136.35 tok/s

    Decode

    504.8 tok/s

    Prefill

    0.59 s

    TTFT

    13.82 GB

    VRAM

    —

    MTP

    391.8 W

    Power draw

    L2 batch seed 2026-10-22: llama.cpp on hw_id=1 running model_id=95 at IQ3_XS