Meta Platforms, the parent company of Facebook, is reportedly set to launch its latest artificial intelligence (AI) model, LLaMA 4, later this month. The release of LLaMA 4 comes a year after the introduction of LLaMA 3 in April of last year.
The IT news outlet D Information reported on the 5th (local time) that "Meta has previously delayed the launch of LLaMA 4 twice, so there remains a possibility of another delay this time." The report stated that "one of the main reasons for the release delay has been that the LLaMA 4 development process has not met expectations on technical standards, particularly in reasoning and mathematical tasks."
Meta plans to invest $65 billion this year to build its AI infrastructure. D Information noted that LLaMA 4 is expected to utilize certain technical elements of machine learning called the 'mixed expert approach' from DeepSeek. This is a method of training parts of the model separately for specific tasks to make it an expert in that field.
Meta is reportedly considering releasing LLaMA 4 first through its chatbot "Meta AI" and later making it open-source. LLaMA 3, released last year, can converse in eight languages, generate higher-quality text, and solve more complex math problems than its predecessor.