Meta's latest open source AI models challenge GPT, Gemini, and Claude
Apr 06, 2025 - digitaltrends.com
Meta has announced Llama 4, the latest iteration of its open-source AI model family, which includes four models: Llama 4 Scout, Llama 4 Maverick, and Llama 4 Behemoth. The models are trained on large amounts of unlabeled text, image, and video data, indicating multimodal capabilities. Currently, Llama 4 Scout and Llama 4 Maverick are available on Meta’s platforms like WhatsApp, Messenger, and Instagram Direct, as well as open-source repositories such as Hugging Face. Llama 4 Behemoth is still in training and is expected to outperform comparable models, serving as a teacher for the other models. Meta has compared Llama 4 models to competitor AI technologies, finding that Llama 4 Maverick excels in creative writing and outperforms models like OpenAI GPT-4o and Google Gemini 2.0 in coding, reasoning, multilingual, long-context, and image generation.
Meta is facing competition from Chinese AI company DeepSeek, known for its cost-effective models, prompting Meta to study their development methods. The Llama 4 Scout model can run on one Nvidia H100 GPU, while the Maverick model requires one Nvidia H100 DGX graphics system. Meta plans to host its first LlamaCon AI conference on April 29 and launch a standalone Meta AI chatbot in the second quarter. Meanwhile, OpenAI has delayed the launch of its GPT-5 model, with CEO Sam Altman announcing upcoming o3 and o4-mini reasoning models as alternatives, with GPT-5 expected to launch in the coming months.
Key takeaways:
Meta has announced Llama 4, an open-source AI model family, with models like Llama 4 Scout and Llama 4 Maverick available for use, while Llama 4 Behemoth is still in training.
Llama 4 Maverick outperformed OpenAI GPT-4o and Google Gemini 2.0 in various functions but struggled against newer models like Gemini 2.5 Pro and GPT-4.5.
Meta is studying the cost-effective AI models of Chinese company DeepSeek to reduce hardware costs for training its powerful models.
Meta plans to host its first LlamaCon AI conference on April 29 and launch a standalone Meta AI chatbot in the second quarter of the year.
