Meta recently announced the launch of the latest member of its Llama series—Llama3.370B. Ahmad Al-Dahle, Meta's Vice President of Generative AI, shared this news on the X platform, noting that Llama3.370B has made significant improvements in performance compared to Meta's previous largest Llama model, Llama3.1405B, while also dramatically reducing costs.
Al-Dahle stated that by leveraging the latest post-training technologies, Llama3.370B not only enhances performance but also lowers operational costs. According to benchmark results released by Meta, Llama3.370B outperformed Google's Gemini1.5Pro, OpenAI's GPT-4, and Amazon's newly released Nova Pro in various areas, particularly excelling in the MMLU test, which evaluates a model's language understanding capabilities.
The model is now available for download from sources such as Hugging Face and the official Llama platform. Meta's initiative aims to dominate the AI field through "open" models. The Llama model can be applied in various scenarios and supports commercialization, although Meta has set usage restrictions for some developers, requiring special permissions for platforms with over 700 million monthly users. Nevertheless, the download count for the Llama model has surpassed 650 million, indicating its widespread popularity among AI developers globally.
To support the future training of larger-scale AI models, Meta is heavily investing in computing infrastructure. The company recently announced plans to build a $10 billion AI data center in Louisiana, which will be Meta's largest AI data center to date. Zuckerberg mentioned in the earnings call that the computing power required to train the next-generation Llama4 model will be ten times that of Llama3. Meta has already procured over 100,000 Nvidia GPU clusters, matching the resources of competitors like xAI.
As the costs of training generative AI models continue to rise, Meta's capital expenditures are also on the rise, with a nearly 33% increase in capital spending in the second quarter of 2024, reaching $8.5 billion. This growth is primarily due to Meta's ongoing investments in servers, data centers, and network infrastructure.
