Zuckerberg will launch a new AI model Llama 4 this month, will compete with ChatGPT

Meta Platforms is set to release its new large language model Llama 4 by the end of this month. The release of this model has already been delayed at least two times, a report by The Information states. The company owned by Facebook is attempting to get a headstart on the AI race.

The report also indicated that Meta could again delay the launch of Llama 4. The information has been provided referencing two close sources. After the success of OpenAI's ChatGPT, large technology companies are pouring a lot of money into AI infrastructure. It has created a massive shift in the technology sector and has provided a new boost to investment in machine learning.

According to the report, during the development of Llama 4, it could not meet Meta's technical expectations, especially in reasoning and mathematical tasks. Apart from this, the company was also concerned that Llama 4's ability to converse in human voice is weak compared to OpenAI's models. Meta plans to spend about $65 billion (about Rs 5,39,000 crore) this year to expand AI infrastructure. There is pressure from investors that big tech companies should show concrete results on their investments.

At the same time, the emergence of a cheap but popular model from Chinese tech company DeepSeek is also breaking the notion that billions of dollars are needed to develop the best AI model. The report also says that Llama 4 could borrow some technical features from DeepSeek. One version of it will use a machine learning technique called "mixture of experts", in which different parts of the model are trained on different tasks so that they become experts in that area.