1 week Ago By Priya Singh
Meta has officially launched its next-generation AI models after months of reported delays. The company introduced three new variants—Llama 4 Maverick, Llama 4 Scout, and Llama 4 Behemoth—though only the first two are currently available to users. These models represent Meta's latest push in the generative AI space and are designed to enhance performance across a variety of tasks, including coding, reasoning, and multilingual processing.
Never a dull day in the AI world! Congrats to the Llama 4 team, Onwards!— Sundar Pichai (@sundarpichai) April 5, 2025 Shortly after the announcement, Google CEO Sundar Pichai responded with a congratulatory message on X, saying, “Never a dull day in the AI world! Congrats to the Llama 4 team, Onwards!” This gesture comes despite Meta’s models directly competing with Google’s Gemini lineup. Meta claims Llama 4 Maverick outperforms Gemini 2.0 Flash in multiple benchmarks, while Llama 4 Scout reportedly delivers better performance than Gemini 2.
0 Flash Lite and Google’s Gemma 3. ALSO SEE: Indian Govt Issues High-Risk Alert For Apple Users; Advices To Update Now: Check Details Meta's forthcoming models, including Llama Behemoth and Llama Reasoning, are also expected to go head-to-head with Google's top-tier AI offerings. Meta CEO Mark Zuckerberg has laid out a bold vision for the company’s latest AI initiative, stating in an Instagram video that Meta’s goal is to build the world’s leading AI, open source it, and make it universally accessible.
Zuckerberg emphasized his long-standing belief that open-source AI will eventually dominate the field, and with the release of Llama 4, he believes that shift is beginning to materialize. Details and pricing here - https://t.co/JGyfhFIYAe — Sundar Pichai (@sundarpichai) April 4, 2025 Among the new models, Llama 4 Maverick features 17 billion active parameters and 128 experts, positioning it as a versatile "workhorse" ideal for tasks like image understanding and creative writing.
On the other hand, Llama 4 Scout, with 17 billion active parameters, 16 experts, and 109 billion total parameters, is optimized for document summarization and reasoning over complex code bases—making it a powerful tool for more technical and analytical applications..
