Arao AI
Back to search results

Llama-3.2-3B-Instruct · FP16

Start a discussion
NVIDIA RTX 309024GBStrata v0.50L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (FP16) on NVIDIA RTX 3090 with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

145.3 tok/s

Decode

Decode speed

693.51 tok/s

Prefill

Prefill speed

0.85 s

TTFT

Time to first token

13.76 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · Q2_K Decode 268 · Prefill 1209.31 ·
  2. Strata · FP16 (current)Decode 145.3 · Prefill 693.51 ·
  3. llama.cpp · Q8_0 Decode 228.2 · Prefill 1711.5 ·
  4. llama.cpp · Q5_K_M Decode 331.9 · Prefill 2489.3 ·
  5. llama.cpp · Q4_K_M Decode 405.6 · Prefill 3042 ·

Core figures

Decode (avg)
145.3 tok/s
Prefill (avg)
693.51 tok/s
TTFT (avg)
0.85 s
VRAM (avg)
13.76 GB
MTP acceptance rate
0.55
TTFB
1.16 GB
Power draw
300.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.2-3B-Instruct
Quantization
FP16
Framework
Strata
Version
v0.50
Context length
32768 tokens
Batch size
1024
GPU layers
15
Flash Attention
On

Hardware

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

GPU
NVIDIA RTX 3090
Nominal VRAM
24 GB
Measured VRAM (avg)
13.76 GB
OS
Ubuntu 24.04
Driver
565.57
CUDA
13.0
Power draw
300.5 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

    v0.50 · Ubuntu 24.04 · CUDA 13.0 · 32768 ctx

    145.3 tok/s

    Decode

    693.51 tok/s

    Prefill

    0.85 s

    TTFT

    13.76 GB

    VRAM

    0.55

    MTP

    300.5 W

    Power draw

    L2 batch seed 2026-10-18: Strata on hw_id=2 running model_id=70 at FP16