Meta Platforms has announced the release of its newest series of large language models (LLM), LLaMA 4, marking a significant advancement in the field of artificial intelligence. The release includes LLaMA 4 Scout, LLaMA 4 Maverick, and a preview of LLaMA 4 Behemoth, each designed to deliver more powerful and personalized multimodal experiences.

LLaMA 4 Scout and Maverick: Leading the Multimodal Revolution

LLaMA 4 Scout is equipped with 17 billion active parameters and 16 experts, making it the world's best multimodal model in its class. It stands out for its efficiency, fitting into a single NVIDIA H100 GPU, and offers an industry-leading context window of 10 million tokens. This model surpasses previous versions and competitors, achieving superior results across a wide range of benchmarks.

LLaMA 4 Maverick, also a 17 billion active parameter model with 128 experts, excels in multimodal performance, outperforming industry benchmarks such as GPT-4.0 and Gemini 2.0 Flash. It offers a remarkable performance-to-cost ratio and an experimental chat version with an ELO score of 1417 on LMArena, demonstrating its proficiency in reasoning and coding tasks.

LLaMA 4 Behemoth: A Glimpse into the Future

The LLaMA 4 Behemoth, a model with 288 billion active parameters and 16 experts, remains under training. Despite this, early performance indicators show it surpassing GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on crucial STEM benchmarks. LLaMA 4 Behemoth's advanced capabilities make it one of the smartest LLMs globally, setting a new precedence in AI innovation.

Accessible AI for All

As part of Meta's commitment to openness and innovation, LLaMA 4 Scout and Maverick are available for download on llama.com and Hugging Face. Additionally, users can experience LLaMA 4 through Meta AI in platforms like WhatsApp, Messenger, and Instagram Direct. Meta aims to ensure that the leading models and systems are transparently available to foster a new era of personalized AI-driven experiences.

Future Prospects

Meta's introduction of the LLaMA 4 series signifies the beginning of a transformative phase for the LLaMA ecosystem. The company remains dedicated to advancing AI research, promoting openness, and providing developers and enterprises with the tools necessary to integrate cutting-edge AI capabilities into their products and services.

Join Meta at LlamaCon on April 29 to learn more about the future of AI and the exciting developments within the LLaMA ecosystem.