Meta has just announced the release of Llama 3.1, the latest version of its open-source artificial intelligence model, which comes in three different versions. This new iteration is the most advanced AI model from Meta to date, underscoring the social network’s significant investment in AI to keep pace with leading startups like OpenAI and tech giants such as Google and Amazon.
The unveiling of Llama 3.1 also highlights Meta's deepening partnership with Nvidia, a key provider of the GPUs used to train Meta’s AI models. Unlike competitors who monetize their AI models, Meta is committed to keeping Llama open source, allowing free access to the technology. This strategic move is designed to attract top talent and reduce computing costs while fostering innovation.
Meta will continue its approach of collaborating with other tech companies to make Llama 3.1 accessible via their cloud platforms. Partners include Amazon Web Services, Google Cloud, Microsoft Azure, Databricks, and Dell. Although these partnerships generate some revenue, Meta emphasizes the broader benefits of open-source accessibility.
The launch comes ahead of a major conference on advanced computer graphics, where Meta CEO Mark Zuckerberg and Nvidia CEO Jensen Huang will speak. Meta, a significant Nvidia customer, relies on the latest GPU technology to train its AI models for various internal uses, including targeting and product development.
For Nvidia, Meta's use of open-source models can drive demand for its chips, as other companies adopt and adapt these models without licensing fees. This symbiotic relationship benefits both companies, promoting the development of AI applications with Meta’s software at the core.
Meta's largest new model, Llama 3.1 405B, boasts 405 billion parameters, allowing it to perform complex tasks and process large data volumes. Smaller versions, Llama 3.1 8B and 70B, are also available for applications like chatbots and coding assistants.
Meta is showcasing Llama 3.1’s capabilities through its digital assistant on WhatsApp and the Meta.AI website in the U.S. Users can toggle between the powerful 405B model and smaller versions to answer complex math problems or coding queries.
This latest release marks a significant step in Meta's AI journey, emphasizing open-source innovation and strategic partnerships to push the boundaries of what artificial intelligence can achieve.
