Meta has introduced Llama 4, a new suite of AI models that aim to outperform those from OpenAI and Google across various benchmarks. The models include Llama 4 Scout, optimized to fit on a single Nvidia H100 GPU, and Llama 4 Maverick, which competes with top models like GPT-4o. Meta asserts these models, particularly Llama 4 Behemoth with 288 billion active parameters, offer superior performance on STEM benchmarks compared to competitors. The models utilize a 'mixture of experts' architecture, aimed at improving efficiency.
Meta's new Llama 4 models are claimed to outperform OpenAI and Google's offerings across numerous benchmarks, indicating a significant advancement in AI technology.
The new models, Llama 4 Scout and Llama 4 Maverick, are capable of exceeding the performance of major competitors while utilizing fewer parameters and lower resources.
Collection
[
|
...
]