Meta aims to position itself ahead of competitors like ChatGPT and Gemini by developing models that require less computing power while performing complex tasks. More details about the Llama 4 family, including a base model named Behemoth and a reasoning model, are expected soon. These developments will likely be discussed at LlamaCon, Meta's first annual AI developers conference.
Key takeaways:
- Meta has unveiled its latest family of generative AI models, Llama 4, which includes two models named Scout and Maverick.
- Scout is a smaller model designed to run on a single Nvidia H100 GPU with 17 billion parameters and 16 experts, while Maverick is a midsized model with 17 billion parameters and 128 experts.
- The Llama 4 models are open-weights and multimodal, allowing them to generate text, images, and code, and providing developers insight into their construction.
- More information on the Llama 4 family, including a base model named Behemoth and a reasoning model, is expected to be revealed at LlamaCon, Meta's first annual AI developers conference.