Arao AI
Back to search results

Llama-3.2-3B-Instruct · FP16

Start a discussion
NVIDIA RTX 4070 Ti12GBllama.cpp v0.23L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (FP16) on NVIDIA RTX 4070 Ti with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

157.49 tok/s

Decode

Decode speed

748.12 tok/s

Prefill

Prefill speed

1.55 s

TTFT

Time to first token

11.4 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · FP16 (current)Decode 157.49 · Prefill 748.12 ·
  2. Strata · IQ2_M Decode 347.68 · Prefill 1284.54 ·
  3. Strata · Q4_K_M Decode 284.07 · Prefill 1145.96 ·

Core figures

Decode (avg)
157.49 tok/s
Prefill (avg)
748.12 tok/s
TTFT (avg)
1.55 s
VRAM (avg)
11.4 GB
MTP acceptance rate
—
TTFB
1.26 GB
Power draw
288.6 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.2-3B-Instruct
Quantization
FP16
Framework
llama.cpp
Version
v0.23
Context length
2048 tokens
Batch size
256
GPU layers
20
Flash Attention
On

Hardware

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

GPU
NVIDIA RTX 4070 Ti
Nominal VRAM
12 GB
Measured VRAM (avg)
11.4 GB
OS
Windows 11 23H2
Driver
570.86
CUDA
12.6
Power draw
288.6 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

    v0.23 · Windows 11 23H2 · CUDA 12.6 · 2048 ctx

    157.49 tok/s

    Decode

    748.12 tok/s

    Prefill

    1.55 s

    TTFT

    11.4 GB

    VRAM

    —

    MTP

    288.6 W

    Power draw

    L2 batch seed 2026-10-18: llama.cpp on hw_id=42 running model_id=70 at FP16