Arao AI
Back to search results

Llama-3.2-3B-Instruct · IQ2_XS

Start a discussion
NVIDIA RTX 409024GBStrata v0.21L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (IQ2_XS) on NVIDIA RTX 4090 with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

281.21 tok/s

Decode

Decode speed

1,389.98 tok/s

Prefill

Prefill speed

0.16 s

TTFT

Time to first token

5.12 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. vLLM · IQ3_S Decode 362.35 · Prefill 1574.76 ·
  2. Strata · IQ1_S Decode 284.1 · Prefill 1178.93 ·
  3. Strata · IQ2_XS (current)Decode 281.21 · Prefill 1389.98 ·
  4. llama.cpp · Q8_0 Decode 245.7 · Prefill 1842.8 ·
  5. llama.cpp · Q5_K_M Decode 357.4 · Prefill 2680.5 ·
  6. llama.cpp · Q4_K_M Decode 436.8 · Prefill 3276 ·

Core figures

Decode (avg)
281.21 tok/s
Prefill (avg)
1,389.98 tok/s
TTFT (avg)
0.16 s
VRAM (avg)
5.12 GB
MTP acceptance rate
0.55
TTFB
2.25 GB
Power draw
333 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.2-3B-Instruct
Quantization
IQ2_XS
Framework
Strata
Version
v0.21
Context length
2048 tokens
Batch size
512
GPU layers
5
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)
5.12 GB
OS
Windows 11 23H2
Driver
565.57
CUDA
13.0
Power draw
333 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-18

    v0.21 · Windows 11 23H2 · CUDA 13.0 · 2048 ctx

    281.21 tok/s

    Decode

    1,389.98 tok/s

    Prefill

    0.16 s

    TTFT

    5.12 GB

    VRAM

    0.55

    MTP

    333 W

    Power draw

    L2 batch seed 2026-10-18: Strata on hw_id=1 running model_id=70 at IQ2_XS