Metas Latest AI Models Challenge GPT, Gemini, and Claude – As pressure in the generative artificial intelligence market heats up, Meta has unveiled Llama 4, the newest member of its open-source artificial intelligence family.
Metas Latest AI Models Challenge GPT
A total of four models makes up the new AI family; Meta described the Llama 4 Scout, Maverick, and Behemoth in depth. On its artificial intelligence website, Meta explained that “large amounts of unstructured text, image, as well as video data” were used to train the models. This means the models will be able to handle multiple modalities in different ways.
Two of the models, Llama 4 Scout including Llama 4 Maverick, are now accessible to users on Meta’s platforms, such as WhatsApp, Messenger, Instagram Direct, and their artificial intelligence website, Llama.com. This was on Saturday. Hugging Face and other open-source repositories provide developers with access to AI models. The Llama 4 Behemoth system is still being trained and will not be published until it is ready. The Behemoth model will instruct the other members of the Llama 4 family and, according to the corporation, will outperform similar models.
In order to find out how well the Llama four versions worked and in what situations they were most useful, Meta compared them against other artificial intelligence systems. Llama 4 Maverick was recommended by the company as the ideal writing tool for creative projects. Furthermore, on tasks like reasoning, coding, multilingualism, long-context, and picture generation, it surpassed both the OpenAI GPT-4o along with the Google Gemini 2.0 versions. A number of AI models, such as Anthropic Claude 3.7 Sonnet, Gemini 2.5 Pro, as well as GPT-4.5, outperformed Maverick.
Even though Meta asserts that Behemoth can surpass the majority of these models (with the exception of Gemini 2.5 Pro), the business has failed to decrease the hardware expenses associated with training its most potent model. Metas Latest AI Models Challenge GPT
Meta was placed on notice by the Chinese AI developer Deep Seek, according to TechCrunch, due to the attention it earned from its highly competitive yet affordable models. Reportedly, the business has been looking closely at the rival’s model development processes in order to learn how to create significant models, such as the R1 and V3, with lower operational expenses than previous Llama models.
The business went into detail about how one Nvidia H100 GPU can power the Llama 4 Scout version. It only takes one Nvidia H100 DGX video card to run its Llama 4 Maverick model. Metas Latest AI Models Challenge GPT
On April 29, Meta will hold its inaugural LlamaCon AI conference. In addition, CNBC reports that the company’s independent Meta artificial intelligence chatbot will debut in the second quarters of this year. Metas Latest AI Models Challenge GPT
When it comes to the release dates of its main AI models, Meta isn’t the only business being cagey. OpenAI CEO Sam Altman recently said on social networks that fans should anticipate new o3 as well as o4-mini reason models to release in the upcoming weeks as a replacement to GPT-5, which was the model that was originally scheduled to be launched by the firm. This will provide OpenAI more time to get its system up to par, as the CEO explained that GPT-5 will be ready to deploy in the coming months.
