Arao AI
Back to search results

Llama-3-8B-Instruct · IQ3_S

Start a discussion
NVIDIA H100 80GB HBM380GBllama.cpp v0.43L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3-8B-Instruct (IQ3_S) on NVIDIA H100 80GB HBM3 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

289.19 tok/s

Decode

Decode speed

992.46 tok/s

Prefill

Prefill speed

1.19 s

TTFT

Time to first token

13.19 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

17 days ago

Last verified

Performance

  1. llama.cpp · IQ3_S (current)Decode 289.19 · Prefill 992.46 ·
  2. llama.cpp · IQ4_XS Decode 283.18 · Prefill 981.02 ·
  3. vLLM · Q2_K Decode 349.23 · Prefill 1726.15 ·

Core figures

Decode (avg)
289.19 tok/s
Prefill (avg)
992.46 tok/s
TTFT (avg)
1.19 s
VRAM (avg)
13.19 GB
MTP acceptance rate
—
TTFB
2.31 GB
Power draw
639.2 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3-8B-Instruct
Quantization
IQ3_S
Framework
llama.cpp
Version
v0.43
Context length
4096 tokens
Batch size
256
GPU layers
99
Flash Attention
Off

Hardware

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

GPU
NVIDIA H100 80GB HBM3
Nominal VRAM
80 GB
Measured VRAM (avg)
13.19 GB
OS
Ubuntu 22.04
Driver
572.83
CUDA
12.8
Power draw
639.2 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.43 · Ubuntu 22.04 · CUDA 12.8 · 4096 ctx

    289.19 tok/s

    Decode

    992.46 tok/s

    Prefill

    1.19 s

    TTFT

    13.19 GB

    VRAM

    —

    MTP

    639.2 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=55 running model_id=69 at IQ3_S