Meta has been talking about releasing a well-constructed open source AI model since April, and finally released its latest AI model on July 23rd. Meta Llama 3.1 is the company's first model to be released for the AI industry. According to the company, this model is the largest open source AI model to date and has already outperformed top-ranked competitors. In a blog post, Meta said that the latest model outperformed GPT 4o and Anthropic's Claude 3.5 Sonnet in several benchmarks. Last year, Llama 2 could only be compared to the previous generation models, which were at the cutting edge. This year, Llama 3 competes with state-of-the-art models and takes the lead in some areas.
What is Llama 3.1?
Llama 3.1 is Meta's most complex AI model to date. It's worth noting that recent Meta AI releases are based on Llama 3 and are pretty fun to use. In a blog post, Meta CEO Mark Zuckerberg said that just as Linux has surpassed proprietary operating systems, open source AI models are rapidly advancing and poised to surpass proprietary models, reflecting the trajectory of Linux becoming the dominant force powering modern devices and infrastructure.
He draws parallels between Meta's investments in open source AI and the earlier Open Compute Project, which leveraged collaborative innovation to achieve significant cost savings. He expects to see similar results with AI, seeing open source development drive progress and efficiencies, mirroring the success of the Open Compute Project in data center design and operations.
“We believe the release of Llama 3.1 marks a turning point in the industry where most developers will start using primarily open source, and we expect to see this approach continue to grow,” he said.
The models will be available on all major clouds including AWS, Azure, Google, Oracle, etc. Companies like Scale.AI, Dell, Deloitte, etc. are ready to help companies adopt Llama and train custom models using their own data.
It's not surprising that Meta has said nothing about training the model, and the actual source data they used to train Llama 3.1 has not been revealed.
Meta AI now supports multiple languages
Meta is expanding the reach of its Llama-based AI assistant to more countries and languages, and is also introducing a new feature that can generate images based on a person's unique appearance. Meta also announced that MetaAI, the Meta app's assistant, is now multilingual. Meta AI is now available in 22 countries, the latest being Argentina, Chile, Colombia, Ecuador, Mexico, Peru and Cameroon.
Users can also interact with Meta AI through WhatsApp, Instagram, Messenger and Facebook in new languages including French, German, Hindi, Hindi Roman, Italian, Portuguese and Spanish, with more on the way.
