The technology company Meta already has a collection of social media platforms. It has also taken steps into the world of large language models. Meta’s new Llama 4 models are LLMs designed to be open-source. The Llama 4 models were released on Saturday, April 5, 2025. They are called Llama 4 Maverick and Llama 4 Scout. Those interested can try out these models, and they are accessible on several Meta platforms.
Still to come is one more release: that of Llama 4 Behemoth. This upcoming addition is touted as even more impressive than Maverick and Scout. In a way, the efforts of Behemoth are already available for people to experience. Behemoth has already been teaching the two LLama 4 models that are currently available. Behemoth is currently still in the training process. Once training is complete, this model will also be released.
Recent Llama 4 Release
Meta officially introduced the LLama 4 models Scout and Maverick on Saturday, April 5, 2025. According to CNBC, Llama 4 is Meta’s most recent venture into “open-source artificial intelligence software.” Mark Zuckerberg is the CEO of Meta, and he is working “to build the world’s leading AI” and “open source it.” Meta is making its AI innovations “universally available” so that “everyone in the world benefits.”
It is now possible for people to try out the newly released models. These models can be found on Meta platforms such as Instagram Direct, WhatsApp, and Messenger. The Meta AI website also has the models available. CNBC states that “Meta will host its first LLamaCon AI conference on April 29.”
Information About an Upcoming Release
Meta is providing preview information about Llama 4 Behemoth. Behemoth is not yet ready for release, but this model is already impressive. It has already been helping the other models learn. The Meta website describes Behemoth as one of the world’s most intelligent LLMs. LLMs are large language models which is a form of generative AI. Large learning models are designed to process language. An LLM is trained with care to comprehend language and to work with language effectively.
Behemoth has recently been compared with similar LLM products. According to the Meta website, the products were compared according to “several STEM benchmarks.” Meta states that Behemoth “outperforms GPT-4.5, Claude Sonnet 3.5, and Gemini 2.o Pro.”
Details About the Llama 4 Models
According to ProjectPro, “LLM parameters help… models grasp the subtlety of language.” These parameters are variables presented to a large language model as the model is trained. They help an LLM utilize language with accuracy and versatility. The linguistic effectiveness of an LLM is impacted by the number of parameters it has and the precision. Effective parameters can provide an LLM with the ability to offer relevant and creative responses.
The Meta LLama 4 models contain many of these parameters that facilitate language versatility. The models are designed to “enable people to build more personalized experiences.” According to Meta, Scout is “a 17 billion active parameter model.” Maverick has the same number of active parameters. Behemoth has many more parameters than the other two models. It is “a 288 billion active parameter model.”
Possibilities for These Large Learning Models
Overall, these large learning models are designed to be applicable in a variety of settings. Meta wants to make sure that the LLama 4 models “are capable of generalized action.” Meta also wants to ensure that the models can have discussions with people with ease. In addition, Meta prioritizes the ability of the models to productively take on new challenges.
Meta’s large learning model software is also intentionally open-source. The Meta website states that the company works under the premise that “openness drives innovation.” Meta wants everyone to be able to contribute to the “future of personalized experiences.” The goal of developing these LLMs is to facilitate many possibilities for those experiences.
