Meta Launches Llama 3 Model; Integrates AI Across Meta Apps

meta ai llama 3 model introduced
Image Courtesy: Meta
In Short
  • Meta has introduced Llama 3 models in two sizes: 8B and 70B parameters. They are open-source and have a context window of 8K tokens.
  • Meta is also training its largest multimodal model with 400B parameters, but it's not available currently.
  • The social media giant is rolling out Meta AI experiences across all its apps including WhatsApp, Facebook, Messenger, Instagram, and more.

After almost nine months, Meta has introduced a new family of Llama 3 models. And they are open-source, just like before, under Meta’s Community License Agreement. Currently, Meta has announced Llama 3 models in two sizes: Llama 3 with 8B parameters and Llama 3 with 70B parameters. They have a context length of 8K tokens.

Meta is also training its largest multimodal model with 400 billion parameters, and it will be available in the near future. However, we are not sure if Meta will choose to open-source their largest model when they release it to the public.

As of now, Meta is integrating its Llama 3 models across all its social media apps including Facebook, Instagram, WhatsApp, Messenger, and on the web as well. Yes, just like ChatGPT, you can now chat with Llama 3 models via meta.ai (visit) for free. And on apps, you can find Meta AI at work in your feed, chats, and search.

meta ai experience in whatsapp
Image Courtesy: Meta

You can use Meta AI to generate instant images and GIFs, find current information on any subject (uses both Google and Bing), ask questions from your feed, and more. You can read our experience of using Meta AI on WhatsApp.

Meta is rolling out AI experiences in its social media apps and launching the web portal in English in several regions: Australia, Canada, Ghana, Jamaica, Malawi, New Zealand, Nigeria, Pakistan, Singapore, South Africa, Uganda, Zambia, and Zimbabwe. The company says more countries will be added in the coming days.

meta llama 3 vs gemini, mistral, gemma, and claude 3 sonnet
Image Courtesy: Meta

Coming to the Llama 3 models, Meta says its 8B model is better than Google’s Gemma 7B and Mistral 7B models. And the Llama 3 70B model beats Gemini 1.5 Pro by a whisker and also scores better than Claude 3 Sonnet in the MMLU benchmark.

We will have to test the models extensively to evaluate where they stand against competing models. But for now, it seems like Meta has launched capable models with the Llama 3 family. What do you think about Meta’s new models? Let us know in the comment section below.

VIA Meta
comment Comments 0
Leave a Reply