Meta launches Llama 3 as most powerful open model

On April 18, 2024, Meta launched Llama 3, at the time the most powerful freely available language model. The initial release included versions of 8B and 70B parameters.

Trained on 15 trillion tokens

Llama 3 was trained on seven times as much data as Llama 2 and performed strongly on reasoning and code. It immediately became the engine behind the Meta AI assistant.

Open source as strategy

With Llama 3, Meta confirmed its course of freely available models, as a counterweight to the closed models from OpenAI and Google.