MetaAI Growth Hits 600M Users with Mark Zuckerberg’s Llama 3.3 Plan

MetaAI Growth Hits 600M Users with Mark Zuckerberg’s Llama 3.3 Plan

MetaAI is getting close to a big goal: 600 million monthly users. CEO Mark Zuckerberg is announcing updates. The release of Llama 3.3, which aims to save time and money, aligns with growth.

Meta said that its AI assistant, MetaAI, is approaching a major milestone, with almost 600 million users every month. In a recent statement, CEO Mark Zuckerberg told everyone about this update.

Meta AI first launched in the fall of 2017 and had more than 500 million users by October. This increase in users comes at the same time that Meta’s newest text model, Llama 3.3, launches. The goal of Llama 3.3 is to enhance performance while reducing operational costs.

Llama 3.3 added a new 70 billion parameter model. Meta says it performs the same as its previous 405 billion parameter model while being cheaper to run. In a performance comparison, Ahmad Al-Dahle, Vice President of Generative AI at Meta, said that Llama 3.3 did better on several benchmarks than Google’s Gemini Pro 1.5 and OpenAI’s GPT-4o.

Improved Performance with Online Preference Optimization

The model uses newer post-training methods, such as online preference optimization, which improves its effectiveness.

Zuckerberg showed what Meta’s AI could become in the future by saying that Llama 4 is coming soon. He said that version 3.3 would be the “last big AI update of the year,” which was a hint at what was to come.

There aren’t many details about Llama 4, but Zuckerberg said earlier this year that it was being trained with more than 100,000 H100 GPUs and that some “smaller” models would be available early next year.

Leave a Reply

Your email address will not be published. Required fields are marked *

NVIDIA Teams with Siam.AI Cloud to Build Thailand’s First Sovereign AI Previous post NVIDIA Teams with Siam.AI Cloud to Build Thailand’s First Sovereign AI
OpenAI's Sora Video Model Skips EU and UK: Regional Availability Revealed Next post OpenAI’s Sora Video Model Skips EU and UK: Regional Availability Revealed