Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

History

Meta releases Llama 3

·

Meta publishes Llama 3 in 8B and 70B, with new safety tools.

Two developers place a model block in a server with a toolbox on the table.
Llama 3 arrived with models and safety tools. Illustration: Denkbaar.

On 18 April 2024, Meta releases Llama 3, the next generation of its open language model.

What happened

  • The first two models have 8 billion and 70 billion parameters, each as a pre-trained and an instruction-tuned version.
  • Meta adds safety tools: Llama Guard 2, Code Shield and CyberSec Eval 2.
  • According to Meta, the models come to the big clouds and to Hugging Face.
  • Meta’s own assistant, Meta AI, runs on Llama 3.
Four model blocks form two pairs of different sizes, with conversation boxes by each pair.
Llama 3 appeared in two sizes, each with base and instruction versions. Illustration: Denkbaar.

Why it mattered

Open models come closer to closed models. For people who want to run a model themselves, Llama becomes a default choice.

A worker studies a licence booklet beside a model block and a screen panel.
Read the licence before putting Llama in a product. Illustration: Denkbaar.

What you can do with it now

Meta calls Llama open source. Still, read the licence that Meta ships with it before you put Llama in a product.

Written afterwards on 12 November 2026, from the source below.

Sources

  1. Meta: Introducing Meta Llama 3: The most capable openly available LLM to date

Topics: open-sourcemodelsmeta