Meta just released Llama 3, and it outshines other LLMs in almost every way

Many tech giants are working on their AI models, and Meta is no exception.  As we previously wrote, Llama 3 is expected to come sooner than we think, and this is exactly what happened.

The new version of Llama is finally here, so without further ado, let’s see what’s new.

Llama 3 is finally here, bringing enhanced performance and more

As reported by MSPowerUser, Meta has just released the latest version of its language learning model, bringing improved performance and scalability.

This allows the AI to perform multi-step tasks, lower false refusal rates, improve response alignment, and understand nuances and complex tasks.

Llama 3 was trained on 15T token of data, which is seven times larger than was used for Llama 2. This version supports 8K context length doubling the capacity of Llama 2.

Regarding performance, Llama 3 8B and 70B outshine the competition in almost every regard.

Meta will also make Llama 3 available to everyone by making smaller versions of Llama 3 available as open-source models.

The new version is more secure, and it will utilize Llama Guard 2, Code Shield, and Cybersec Eval 2.

The model is already available in action and it’s powering Meta AI assistant that can be found in Meta’s services such as Facebook, Instagram, and WhatsApp.

Meta also has plans to bring video and image support to Llama 3, but no release date for these features is known.

Llama 3 was supposed to change our perspective on AI, and now that the model is finally here, did Meta deliver on its promise? Let us know in the comments!