Meta launches the Llama 3.3 70B text model and adds almost 100 million new Meta AI users in the last month

Dec. 9 (Portaltic/EP) –

Goal has launched its new Artificial Intelligence (AI) model Flame 3.3 70Bwith which it ensures similar performance to the flagship model Llama 3.1 405B for text-based applications, but with lower costwhile sharing that he already has almost 600 million monthly active users using its Meta AI ‘chatbot’nearly 100 million more than the previous month.

The chatbot Meta AI arrived in April of this year driven by the model of AI Call 3, which includes capabilities to access information in real time and interact with natural language through the company’s services and platforms, such as WhatsApp, Facebook, Instagram and Messenger.

In this context, the company announced in November that its AI-powered chatbot, Meta AI, It already registered more than 500 million users active on the different platforms and formats in which it is available, a figure that reached less than a year after its launch.

Now, Meta has announced that the number of active Meta AI users has increased to almost 600 million users who use the ‘chatbot’ monthly, which translates into practically 100 million more users from November until now.

This has been made known by the Meta CEO Mark Zuckerbergin a post on Threads, where he shared the latest news of the year related to Meta’s AI, among which he highlighted the amount of growth in monthly active users who use his ‘chatbot’.

NEW MODEL LLAMA 3.3 70B

Another of the new features announced is the launch of its new text model Llama 3.3 70B, which is now available to all developers and which offers a similar performance to the Llama 3.1 405B model when it comes to delivering quality results in text-based applications.

However, as highlighted by the company on its websitethis new model is capable of execute your results “at a fraction of the cost” compared to previous models, allowing its AI capabilities to be used more affordably.

As detailed by the Vice President of Generative AI at Meta, Ahmad Al-Dahlein a post on X (former Twitter), this model is “easier and more profitable to run”. This is because it “leverages the latest advances in post-training techniques”, including online preference optimization, which improves performance at a “significantly lower” cost.

Likewise, as shown in a comparative table between the different Llama models and other models in the sector, Llama 3.3 70B It surpasses Google’s Gemini Pro 1.5 AI and OpenAI’s GPT-4o by several points.

Zuckerberg has also stressed that, the Meta’s next AI release will be the Llama 4 modelalthough no information has been provided in this regard. Finally, he has also indicated that the company is building a data center of more than 2 GW in Louisiana (United States), which they plan to use to train future versions of the Llama models, including Llama 4.