Meta is making waves in the AI world with its latest Llama 4 models, and it’s a big deal for the US in the global AI competition. David Sacks, who advises the White House on AI and crypto, believes that the release of Llama 4 has catapulted the US to the forefront of the AI race. Sacks emphasized the significance of open-source technology in this journey, mentioning, “To win the AI race, the US needs to excel in open source too, and Llama 4 is helping us regain the lead.”
His comments come at a time when the US and China are both striving for AI dominance. Since taking on his role during President Trump’s administration, Sacks has been vocal about the necessity for the US to maintain its edge in AI. He’s confident in America’s capabilities but warns against becoming complacent.
Meta recently introduced the fourth generation of its Llama models, dubbed Llama 4 Scout and Llama 4 Maverick. They’re being hailed as “the most advanced models yet and the best in their class for multimodality.”
The Llama 4 Scout model boasts 17 billion active parameters with 16 experts, reportedly outshining competitors like Gemma 3 and Gemini 2.0 Flash-lite in various benchmarks. On the other hand, Llama 4 Maverick, equipped with 128 experts, also features 17 billion active parameters. Meta claims Maverick outperforms GPT-4o and Gemini 2.0 Flash on numerous benchmarks and competes with DeepSeek v3 in reasoning and coding tasks, all while using only half the active parameters.
Back in July 2024, Meta CEO Mark Zuckerberg expressed optimism that by 2025, Llama models would be acknowledged as the most advanced in the industry. The launch of Llama 1 in February 2023 was met with overwhelming interest, with Meta receiving over 100,000 access requests.
