- Meta launches Llama 4 with three variants: Scout, Maverick, and Behemoth, with a multimodal approach and open source.
- The models incorporate Mixture of Experts (MoE) architecture and context windows of up to 10 million tokens.
- Scout and Maverick are already available on apps like WhatsApp and Instagram, while Behemoth is still in development.
- Distribution of Llama 4 is restricted outside the US by legal frameworks, particularly in the EU.
Meta has taken a new step in the evolution of its Artificial Intelligence ecosystem with the launch of Llama 4, a collection of models that seek to compete with current industry giants such as GPT-4 or Gemini. With a revamped architecture and the addition of multimodal capabilities, Mark Zuckerberg's company is trying to strengthen its position in a race marked by the innovation and the speed of development.
The Llama 4 family comes with three main models: Scout, Maverick and Behemoth, each with distinct features designed to cover a wide range of needs, from everyday tasks to highly specialized uses. Furthermore, Meta's commitment remains clear: to maintain the open source as a differentiating element, which places these models in a prominent place within the panorama of accessible tools for member and individual developers. For the latter, the open source Llama 4 can offer the opportunity to experiment and develop custom applications.
A new architecture for a new generation
One of the most relevant aspects of Llama 4 is its architecture based on the Mixture of Experts (MoE) technique.. This approach allows, instead of activating all the model parameters in each query, only the parts necessary to accomplish a given task come into play. This not only improves the efficiency, but also reduces the computational load and accelerates responses. With the advancement in architecture, AI developers can explore new possibilities, such as those presented in the article on Microsoft Loop.
Llama 4 Scout and Maverick share a database of 17.000 billion active parameters., although they differ in the number of experts they use: Scout uses 16, while Maverick uses 128. This configuration differentiates their capabilities, focusing Scout on technical tasks such as code analysis and summary of documentation, already Maverick in more general functions such as the Creative writing and interpretation of text and imageFor those interested in code analysis, the architecture of Llama 4 may be crucial and you can find more information in the post about Python.
Instead, Behemoth represents a commitment to the frontier of model development.. Although he is still in the training phase, it has been revealed that he has 288.000 billion active parameters distributed among 16 experts. According to Meta, its initial tests place it above competitors such as GPT-4.5, Claude 3.7, or Gemini 2.0 Pro in benchmarks focused on specific areas. STEM.
Multimodality as the key to the future
Multimodality is another of the distinctive signs of this new generation. Scout and Maverick are designed to understand and work with different types of inputs such as text, images and even video, thus allowing the development of more versatile and adaptable assistants. This translates into more sophisticated applications within corporate, educational, or creative environments, where the simultaneous interpretation of visual and textual information is essential. In this context, the integration of models like Llama 4 could contribute to improving the capabilities of modern tools, similar to what is observed in Google Colab.
One of the most innovative elements is the introduction of context windows of up to 10 million tokensWith this capability, models can handle extremely large blocks of data, facilitating deep analysis of large documents or complex codebases. This expansion of the context is unprecedented in previous models. and could become a standard for future AI releases globally.
Thanks to the MoE architecture, Llama 4 models also consume fewer resources during training and inference., something essential to scale its use without the need for excessively expensive infrastructure. Meta mentions that Scout can work with a single Nvidia H100 GPU, which democratizes access for small businesses or developers with more limited resources. This resource-efficient approach is similar to that of WiFi connection systems, where resource optimization is key to concrete benefits.
Availability and international deployment
The Scout and Maverick models are now available on several Meta platforms., such as WhatsApp, Instagram, Facebook Messenger, and the Meta AI website. For those who want to get into a more technical way or develop them on their own, versions are available at llama.com and the popular platform hugging faceThis international availability marks a significant advance in access to Artificial Intelligence, as discussed in the article on Microsoft news.
However, not everyone will be able to access these advances equally. Meta has confirmed that both multimodal functions and model distribution are restricted in the European Union. for legal reasons related to the Data Protection and regulations on artificial intelligence. This means that companies located in EU territory cannot, for now, use or redistribute them without a specific license.
In addition, corporations with more than 700 million active users per month They will need to apply for additional permits to implement these technologies. This could slow their deployment in markets like Europe, although Meta has expressed its intention to address these obstacles in future updates.
One step closer to the reasoning model
Beyond the current models, Meta has already publicly announced that it is working on a fourth variant: Llama 4 ReasoningAs its name suggests, it will focus on improving logical reasoning, something that its current versions have yet to fully integrate. The introduction of this model would mark Meta's first attempt to compete in an arena where OpenAI and other companies have already made significant strides.
Currently, Llama 4 models respond more freely to social or political issues, something Meta has adjusted so that new versions are less reticent about addressing controversial topics. This reduction in filters follows the trend seen in recent models like Grok 3 and is part of a broader shift in the industry toward more open and less censored interactions.
Furthermore, Meta has hinted that Llama 4 Behemoth could serve as a basis for the creation of smaller, more specialized models., which would allow its power to be applied in more specific contexts without the need to deploy the entire infrastructure.
The new generation of Llama 4 puts Meta in a strong position in the competition for the development of advanced AI models.With a combination of open source, efficient architecture, and multimodal capabilities, Scout and Maverick already offer solutions applicable today. Meanwhile, Behemoth promises to be a cornerstone for building even more powerful models in the future. Although its distribution is limited by legal issues in some regions, the technical advancement seems indisputable and reinforces Meta's commitment to leading the open AI sector.
