Meta has unveiled its final major AI update of the year, with CEO introducing the highly anticipated 70-billion-parameter Llama 3.3 model.
Meta has unveiled its final major AI update of the year, with CEO Mark Zuckerberg introducing the highly anticipated 70-billion-parameter Llama 3.3 model. This new model promises enhanced efficiency, achieving almost the same performance as Meta’s 405-billion-parameter model, but at a much lower computational cost.
The Llama 3.3 model, which is instruction-tuned and text-only, shows improved performance compared to its predecessors, Llama 3.1 70B and Llama 3.2 90B, in text-focused applications. In certain instances, Llama 3.3 70B approaches the performance levels of Llama 3.1 405B with notable proximity, signifying a substantial enhancement in Meta’s artificial intelligence portfolio.
Meta’s Llama system has significantly increased in prominence and utilization, now being acknowledged as the most extensively adopted AI model globally. Zuckerberg disclosed that the Llama model has been downloaded more than 650 million times, thereby reinforcing its open-source foundation.
