On Tuesday Meta announced the launch of its latest and largest artificial intelligence (AI) model in a blog post. The company claimed that the new open-source model can outperform major closed AI models such as GPT-4, GPT-4o, and Claude 3.5 Sonnet across several benchmarks.
The new model is called Meta Llama 3.1 405B. The newer versions were filtered from the 405B model and now offer a 1,28,000 tokens context window. Meta claims both of these models are now the leading open-source large language models (LLMs) for their sizes.
Meta introduces Meta Llama 3.1 405B AI Model
Llama 3.1 can be defined as the most complex and advanced AI model by Meta till now. CEO Mark Zukerberg of Meta explained that open-source AI models are rapidly upgrading itself with more advancements. There are high chances that this change can overtake the dominant force of Linux in powering modern devices.
It is expected that Llama 3.1 will be available on major cloud platforms. This includes Azure, AWS, Oracle and Google, among others. In addition to this companies such as Deloitte, Dell and Scale.AI are expected to train custom Llama with their own data. This can eventually help other enterprises to adopt the AI model on a large scale.
When Meta announced the 3.0 model earlier in April, they had hinted at a more powerful model launch during Summer. The 405B parameter model is a significant announcement as it enables app developers with advanced capabilities across use cases such as multilingual chatbots, long form text summarization, language translation, code generation and code analysis – to name a few.
“While there are other providers offering open models, Meta continues to release very large parameter models with access to model weights. The gap between the capabilities of closed sourced vs. open models is increasingly narrowing. However, closed source models still have an advantage due to them being offered as a cloud service and the broad integration they have with AI platform tools. Hence, Meta’s partnerships with enterprise AI platform vendors is critical to ensure broader distribution and tighter integration with popular AI platform tools,” Arun Chandrasekaran, VP, Analyst – Tech Innovation, Cloud and AI, Gartner, explained.
Key features of Llama 3.1 405B AI Model
According to an official blog, Meta explained that Llama 3.1 contains a bunch of features that could make it the best AI model. Given below are the key features of Llama 3.1
- Real-time and batch inference
- Supervised fine-tuning
- Evaluation of your model for your specific application
- Continual pre-training
- Retrieval-Augmented Generation (RAG)
- Function calling
- Synthetic data generation
Other updates
The previously released Llama 3 8B and 70B AI models have also been upgraded. Meta will also expand the reach of its Llama-based AI assistant to additional countries and languages. Additionally, it will also introduce a new feature that enables image generation based on a person’s unique appearance. Meta has also announced that the Meta AI, an assistant in Meta apps, is now going multilingual.
Meta AI is now available in 22 countries, with the newest countries including Argentina, Chile, Colombia, Ecuador, Mexico, Peru and Cameroon.
Furthermore, the users can also interact with Meta AI across WhatsApp, Instagram, Messenger and Facebook in new languages. You can now use Meta AI in French, German, Hindi, Hindi-Romanised Script, Italian, Portuguese and Spanish.
Follow FE Tech Bytes on Twitter, Instagram, LinkedIn, Facebook
