Meta’s Llama 4: A Game Changer in the AI Race, Puts the US Back in the Lead, Says David Sacks

Meta has made a bold claim with the release of its latest AI model, Llama 4, asserting that it "outperforms" its competitors in several key areas. This new model is seen by many as a critical step in ensuring that the United States remains a leader in the global artificial intelligence race, with White House AI and crypto czar David Sacks commenting that Llama 4 has put the US back in the lead.

Sacks, who has been an outspoken advocate for US dominance in the AI field, took to X (formerly Twitter) on April 5, stating, “For the US to win the AI race, we have to win in open source too, and Llama 4 puts us back in the lead.” His statement highlights the growing competition between the US and China as both countries vie for supremacy in AI technology.

Source: David Sacks

Since his appointment as the White House AI and crypto czar following President Donald Trump’s inauguration in January, Sacks has been vocal about the need for the US to maintain a competitive edge in artificial intelligence. In his early days on the job, Sacks expressed confidence in the US’s ability to lead, but cautioned against complacency.

Llama 4: Meta’s Most Advanced AI Models Yet

Meta’s latest release, the Llama 4 series, includes two new models: Llama 4 Scout and Llama 4 Maverick. These models are being hailed as the company’s most advanced to date, with the AI division describing them as “the best in their class for multimodality.”

According to Meta, the Llama 4 Scout model features 17 billion active parameters and employs 16 experts. The company claims that this model outperforms competing large language models such as Gemma 3, Gemini 2.0 Flash-lite, and Mistral 3.1 across a broad range of widely accepted benchmarks.

Meanwhile, the Llama 4 Maverick model also boasts 17 billion active parameters, but with a more powerful configuration of 128 experts. Meta asserts that Maverick outperforms other leading models like GPT-4 and Gemini 2.0 Flash, and can perform on par with DeepSeek v3 in reasoning and coding tasks, despite using only half the active parameters.

Meta’s Ambition: Llama Models to Lead in AI by 2025

This announcement comes less than a year after Meta CEO Mark Zuckerberg stated in July 2024 that he expected Llama models to become “the most advanced in the industry” by 2025. Meta’s commitment to advancing its AI capabilities is evident, as the company has rapidly scaled up its efforts since releasing the first limited version of Llama 1 in February 2023.

The demand for Meta’s Llama models has been staggering. In 2023, Meta reported receiving over 100,000 requests for access to Llama 1, a clear sign of the model’s immense appeal. Meta’s continued push to improve its Llama models demonstrates its ambition to be at the forefront of AI innovation, and Llama 4 is a critical part of that vision.

Llama 4 Maverick instruction-tuned benchmarks. Source: Meta

A Growing Race for AI Supremacy

The release of Llama 4 has intensified the global AI race, with major players like Meta, OpenAI, and others striving to lead in developing the next generation of AI technology. As the US looks to retain its edge in the AI sector, Meta’s Llama 4 is seen as a crucial milestone that will influence not just the tech landscape but also the future of AI regulation and innovation worldwide.

With the global spotlight now on AI, the competition between the US and China continues to heat up. As more advancements are made and more countries invest in artificial intelligence, the race for AI supremacy is far from over. However, with Llama 4, Meta has undoubtedly thrown its hat into the ring, pushing the US to the forefront of the AI revolution.