Meta Platforms released the biggest version of its Llama 3 artificial intelligence (AI) models on Tuesday, boasting multilingual skills and general performance metrics that rival the industry’s biggest players like Open AI. Meta will be publicly releasing Meta Llama 3.1 405B, the world’s largest and most capable openly available foundation model.
“Until today, open source large language models have mostly trailed behind their closed counterparts when it comes to capabilities and performance. Now, we’re ushering in a new era with open source leading the way,” commented Meta.
Largest model yet
Meta’s latest Llama 3 rivals top AI models in general knowledge, steerability, math, tool use and multilingual translation. The tech giant expects the latest model to encourage new applications and modeling paradigms including synthetic data generation to enable the improvement and training of smaller models.
The new Llama 3 model can converse in eight languages, write higher-quality computer code, and solve more complex math problems.
As part of this latest release, Meta is introducing upgraded versions of the 8B and 70B Llama 3 models. These are multilingual models that have a significantly longer context length of 128K, state-of-the-art tool use, and overall stronger reasoning capabilities. This enables the tool to support more complex use cases, such as long-form text summarization, multilingual conversational agents, and coding assistants.
Notably, the company trained its largest model yet on over 15 trillion tokens. Compared to previous versions of Llama, Meta improved both the quantity and quality of the data it used for pre-and post-training.
Developing AI responsibly
As part of its ongoing efforts to develop AI responsibly beyond the model layer and help others to do the same, Meta will release a full reference system. This includes several sample applications and includes new components such as Llama Guard 3, a multilingual safety model. In addition, it will include the Prompt Guard, a prompt injection filter.
“Unlike closed models, Llama model weights are available to download. Developers can fully customize the models for their needs and applications, train on new datasets, and conduct additional fine-tuning,” added Meta.
Read: Tesla profits plunge 45 percent in Q2 2024, revenues rise to $25.5 billion
Ensuring greater AI access
Meta added that while many may argue that closed models are more cost-effective, Llama models offer some of the lowest cost per token in the industry, according to testing by Artificial Analysis.
For his part, Mark Zuckerberg noted that open source will ensure that more people around the world have access to the benefits and opportunities of AI. Moreover, it ensures that power isn’t concentrated in the hands of a small few and that the technology can be deployed more evenly and safely across society.
“That’s why we continue to take steps on the path for open-access AI to become the industry standard,” the company added.
For more news on technology, click here.
