Meta, the parent company of Facebook, Instagram, and WhatsApp, has announced it will release a commercial version of its open source large language AI model, LLaMa.

The language model, which stands for Large Language Model Meta AI, was first announced in February and was made available to researchers as a small foundational AI model.

According to the Financial Times, Meta will release a commercial version of the model, allowing developers to build applications using the foundational model.

Commercial access would mean that businesses and developers could adapt and improve the open-source AI model, potentially leading to further innovation and improving the functionality of LLaMA.

Recommended

• Scotland’s Synthetic Data Journey
• US Data Breaches on Track to Reach Record High
• Report: Over Half of UK Companies Overwhelmed by Data

Meta’s AI model is being offered with much smaller parameters compared to ChatGPT – and the more parameters, the more points a model was trained on, and the greater its output will be in terms of quality. LLaMA will be available with 7, 13, 33 and 65 billion parameters, while ChatGPT’s -3.5 model has 175 billion, and GPT-4 has an estimated one trillion.

The advantage of have smaller parameters – such as requiring less computing power and resources – made sense for LLaMA’s use as tool for academics and researchers.

Since it is still a foundational model and is trained on a large set of unlabelled data, it can be useful to developers fine tuning AI to different tasks and features.

Meta’s launch of a commercial AI model means it has officially joined the AI race, competing with OpenAI’s landmark ChatGPT, and Google’s Bard which is currently attempting to catch up.

While Meta’s LLaMA is smaller than the two largest models, it does have larger parameters than other AI models on the market, and it’s open-source availability may soon mean a boom in innovation for smaller-scale models.