Meta’s Llama Hits 1 Billion Downloads, Doubling Growth in Three Months

Meta’s open source AI model family, Llama, has surpassed 1 billion downloads, CEO Mark Zuckerberg announced on Threads. This milestone marks a sharp rise from 650 million downloads in December 2024—a 153% surge in just three months.

First introduced for researchers, Llama has since expanded its reach to developers and AI enthusiasts worldwide. 

View on Threads

Ahmad Al-Dahle, VP, head of GenAI at Meta, posted on LinkedIn about the achievement saying that in the span of two years, Llama went from a research project to the most widely used open source AI model.

“I’m so proud of my team for working tirelessly to make Llama better everyday. Meta’s GenAI org actually turned 2 years old on February 27th and it has been so inspiring and humbling to see them work so hard to improve Llama in such real ways for developers, startups and enterprises around the world,” Dahle said.

He added that this moment is a celebration of the entire open source AI community. “Each Llama download represents a vote of confidence in open source AI – and a shared belief that we can go so much further building together,” he added.

Meta is gearing up to release multiple Llama models in the coming months, including "reasoning" models similar to OpenAI’s o3-mini and versions with built-in multimodal capabilities. Zuckerberg has also hinted at "agentic" features, indicating that some models may have the ability to operate autonomously.

In October last year, Zuckerberg also confirmed AIM’s prediction of the company releasing Llama 4 early 2025 at its third quarterly earnings call. 

“I expect that the smaller Llama 4 models will be ready first, and we expect [them] sometime early next year, and I think that they’re going to be a big deal on several fronts — new modalities, capabilities, stronger reasoning, and much faster,” Zuckerberg said.

The Llama 3 models marked a major turning point in the industry, but Zuckerberg expressed even greater enthusiasm for Llama 4, which he confirmed is well into development. 

“We’re training the Llama 4 models on a cluster that is bigger than 100k H100s or bigger than anything that I’ve seen reported for what others are doing,” he said.

Llama powers Meta AI across Facebook, Instagram, and WhatsApp, and is a key part of the company’s long-term AI strategy.