Tech giant Meta this week released the latest generation of its large language model, Llama 3.1 405B, an open-source model that it says is on par with proprietary LLM competitors like OpenAI. GPT-4 And GPT-4o and Anthropogenic Claude 3.5 Sonnet.
The llama is what feeds the Meta AI As of Tuesday, Llama 3.1 405B is available through the assistant you encounter on WhatsApp (which is owned by Meta) and on the Meta.ai website. While you can also use Meta AI on Instagram and Facebook, it wasn’t immediately clear whether the latest model is also available on those platforms. A spokesperson did not respond to a request for comment.
The first version of Llama de Meta has been freed in February 2023, but even CEO Mark Zuckerberg acknowledged that early versions of Llama lagged behind their peers.
“Last year, Llama 2 was only comparable to an older generation of models across the border,” he wrote. in a blog post published Tuesday.
Large language models are the technology behind generative AI chatbots like OpenAI’s ChatGPT, Google Gemini And Meta AIThey are trained on massive datasets to learn how we use language so they can generate their own unique content that looks at least plausibly human.
In addition to now having access to Llama 3.1 405B, Meta AI’s image-generation feature, Imagine, is starting to enable what the company calls “Imagine Me” prompts, which let you create images of yourself doing things like surfing or as part of a surreal painting, based on existing photos. Meta AI is also getting new editing tools, which will let you remove and edit objects in images. Starting this week, English-speaking users will be able to share these images on Facebook, Instagram, Messenger, and WhatsApp.
Meta AI’s image generator was a feature that impressed my CNET colleague Katelyn Chedraoui in what she otherwise considered a “handy but unimpressive” assistant.
Llama has been downloaded more than 300 million times to date, according to figures from Meta.
Lama against everyone
The latest Llama models, which also include Llama 3.1 8B and 70B, have a context window of 128,000 tokens, which is a measure of how much memory the model can memorize in a given conversation. OpenAI’s GPT-4o and the new GPT-4o Mini also feature 128,000 token pop-ups, while Google’s Gemini 1.5 Pro has a window of 1 million chips.
According to a separate blog postLlama’s enhanced reasoning capabilities help Meta AI understand more complex queries, including math and coding. Meta models also support eight languages.
Like Llama 3, who was released in AprilLlama 3.1 405B was trained on over 15 trillion tokens, which is equivalent to approximately 11.25 trillion words.
Meta claims that the 8B and 70B models are best suited for text summarization and as conversational agents and coding assistants. Meanwhile, the 405B model can be used to create synthetic data, or data generated by algorithms or computer simulations (rather than coming from real sources). It can also be used in model distillation, which is the process of transferring knowledge from one LLM to another. a smaller modelwhich delivers AI capabilities and speed while using fewer computing resources.
More than 25 partners, including Amazon, Databricks, and Nvidia, are launching services associated with Llama 3.1 405B to support these developers, which Zuckerberg says also gives the model a fighting chance.
A key difference between Llama and its peers is that the Meta model is open source. exists in two varietiesProprietary LLMs can only be used by developers who purchase access. Open source LLMs are widely available for free.
Zuckerberg said this would ultimately make Llama and Meta AI more competitive, much like the open-source version of Linux software became more popular than the closed, proprietary versions of Unix software that were developed by big tech companies. In his blog post, Zuckerberg explained that this was because Linux allowed developers to experiment and was more affordable, which led to more users and, ultimately, more progress.
