After reportedly delaying its latest AI range for a few months, Meta unveiled its new Llama models on Saturday. While the company introduced three models: Llama 4 Maverick, Llama 4 Scout and Llama 4 Maverick and Llama 4 Behemoth – only two are currently available.
Soon after the announcement of Meta’s Llama 4 AI models, Meta AI got an unexpected congratulatory message from Google CEO Sundar Pichai who in a post on X wrote, “Never a dull day in the AI world! Congrats to the Llama 4 team, Onwards!”
Notably, Meta’s new AI models compete directly with Google’s offerings. For example, Meta claims that Llama 4 Maverick outperforms Gemini 2.0 (Flash) in coding, reasoning, multilingual, long context and image benchmarks, while Llama 4 Scout is said to deliver better results than Gemini 2.0 Flash Lite and Gemma 3 models.
Additionally, Meta’s upcoming models such as Llama Behemoth and Llama Reasoning will also compete directly with Google’s leading offerings.
All you need to know about Llama 4 models:
Outlining the ambitions for Llama 4 models, Meta CEO Mark Zuckerberg in a video on X said, “Our goal is to build the world’s leading ai, open source it, and make it universally accessible so that everyone in the world benefits. and i’ve said for a while that i think that open source ai is going to become the leading models, and with a lot with llama 4 this is starting to happen.”
Lama 4 Maverick comes with 17 billion active parameters and 128 experts. Meta says Maverick is a “product workhorse model for general assistant and chat use cases” and is ideal for precise image understanding and creative writing tasks.
Meanwhile, Llama 4 Scout has 17 billion active parameters, 16 experts and 109 billion total parameters. The smaller Llama 4 model is said to excel at tasks such as summarising documents and reasoning over code bases.
