Competition in the AI world is intensifying daily. Meta has now entered this race with a powerful move, unveiling its highly anticipated Llama 4 series of AI models, directly challenging ChatGPT and Google Gemini. Launched by Mark Zuckerberg, these models—Llama 4 Scout and Llama 4 Maverick—are integrated into WhatsApp, Instagram, Messenger, and the Meta website.
Significantly, these models are equipped with multi-modal features, enabling them to understand and process text, images, and videos.
A Major Shift in AI Technology
The most notable feature of Meta's new AI models is their image understanding capabilities. They not only read and comprehend text but also accurately process image-related information, setting them apart from many other models. Llama 4 models are trained using a "mixture of experts" machine learning technology, which optimizes different parts of the model for specific tasks. This technology draws inspiration from the Chinese DeepSeek AI startup.
Llama 4 Scout: Outperforming Gemini and Mistral
The Llama 4 Scout model is built with 17 billion active parameters and a total of 109 billion parameters. It incorporates 16 experts, strengthening its multitasking and processing speed. Meta claims this model surpasses larger AI models like Gemini 3, Gemini 2.0 Flashlight, and Mistral 3.1 in benchmark tests. Smaller models are described as highly suitable for tasks such as document summarization, code generation, and logical reasoning.
How to Use Meta's New AI Models?
Meta's new AI models are accessible through platforms like WhatsApp, Instagram, and Messenger. Users can also access these features via the Meta AI website. Currently, multi-modal features are available only to English-speaking users in the United States, but Meta may expand this availability to other countries in the future.
Meta's new offering represents a significant advancement in the AI landscape, potentially posing a b challenge to services like ChatGPT and Gemini in the coming days.
