Meta would offer its open source language model LLaMA as a paid service to companies. Business newspaper Financial Times writes this based on three sources. Until now, LLaMA has been released as a language model for researchers.
The release would take place soon, writes Financial Times. After the release, companies can start creating their own applications based on the large language model, as is already happening with OpenAI’s GPT-3 and GPT-4. The intention is to compete with OpenAI, a source told the business newspaper.
LLaMA was released last winter as an open source language model for researchers. Meta had protected access, but the model leaked. Meta CEO Yann LeCun hinted last weekend on the release of LLaMA during a panel discussion at the Les Rencontres économiques d’Aix-en-Provence conference. “The landscape will change in the coming months or weeks. There will be open source models that are as good as the models that are available now.” He also noted that Meta develops AI models open source. Like other large language models, LLaMA works with large amounts of text as training data, where the model tries to become as good as possible at predicting the next word.