Meta recently announced the launch of its new AI model series, Llama 4, which includes three main models: Llama 4 Scout, Llama 4 Maverick, and Llama 4 Behemoth. These are the most advanced models from the company to date and are designed to enhance AI capabilities across various Meta platforms, including WhatsApp, Messenger, and Instagram.
اضافة اعلان
Llama 4 Details:
Llama 4 Scout
Specifications: A compact model designed to run on a single Nvidia H100 Graphics Processing Unit (GPU), with a context window of up to 10 million tokens.
It outperforms models like Google's Gemma 3 and Mistral 3.1 in a range of metrics, making it ideal for applications that require fast and efficient processing of large amounts of textual data.
Llama 4 Maverick
This larger model compares in performance to models like OpenAI's GPT-4o and DeepSeek-V3 in tasks such as programming and logical thinking, with a lower use of active parameters. It is suited for applications requiring advanced analytical capabilities and code generation.
Llama 4 Behemoth
It contains 288 billion active parameters and a total of 2 trillion parameters. It surpasses models like GPT-4.5 and Claude Sonnet 3.7 in performance. This model is suited for applications that require high computational power and the processing of complex data.
Technical Innovations
Llama 4 models are based on a Mixture of Experts (MoE) architecture, which activates only the necessary parts of the model for each task, improving resource efficiency.
The Llama 4 models are available for download via Meta's site or the Hugging Face platform. They have also been integrated into Meta's smart assistant across platforms like WhatsApp, Messenger, and Instagram.
Meta plans to invest up to $65 billion this year to expand its AI infrastructure, amid growing competition in the field.
Although Meta markets Llama 4 as open-source models, there are restrictions for commercial entities with more than 700 million users, which must seek permission from Meta before using the models. This has drawn criticism from the open-source community.
Meta is expected to discuss its future AI plans at the upcoming LlamaCon conference, scheduled for April 29.
With Llama 4, Meta aims to compete with other models like OpenAI's GPT-4 and Google's Gemini 2.0 Flash, focusing on delivering superior performance in a wide range of tasks.
