Meta announced the launch of Llama 3, the latest version of the main language model (LLM), on April 18, calling it a “huge leap forward from Llama 2.”

The company said that it initially launched the first two models of the current version, which feature 8B and 70B parameters, and the following models are scheduled to feature 400B parameters.

Meta emphasized that Llama 3 was trained on a “large, high-quality training dataset” with more than 15 trillion codes, which is 7 times larger than Llama 2 and features 4 times more code.

Llama 3 also features filtering technologies, including NSFW filters, to ensure data quality.

LLama 3 outperforms Llama 2 and competing models such as Anthropic's Claude Sonnet, Mistral Medium, and OpenAI's Chat GPT-3.5 by more than half the time across 12 use cases.

The first versions of Llama 3 are text models. However, future releases will be multilingual and multimedia. They will also provide a longer context and show better performance in reasoning and coding, which Mita described as “essential LLM skills”.

The company plans to deploy Llama 3 across all major cloud providers, API provider models, and other services. It plans to launch the product “everywhere.”