Meta has announced the latest recurrence of its open-source AI model Family Lama 4, which is developed by the brand, while competition in the generic AI industry continues.
The new AI family consists of four models, and the meta detailed Lama 4 Scouts, Lama 4 Maveral, and Lama 4 Behemth. Meta expanded on its AI website that the model was “trained in large amounts of unbelled text, image and large amounts of video data.” This indicates that the model will have various multimodal capabilities.
Currently, two models- Lama 4 Scouts and Lama 4 are available for Meta platforms for users, including WhatsApp, Messenger and Instagram Direct, as a Saturday, as a Saturday. Developers can use AI models in open-source repository such as Hugging Face. The LLAma 4 behemoth model has not been released yet and is still in training. The company has indicated that the industry should expect to perform better than comparative models from the Bhamoth model, and that this Lama will serve as a teacher for other models in the 4 family.
Between its internal testing, Meta compared Lama 4 models to competitive AI technologies to AI Technologies to determine its ability and best use cases. The company indicated that the Lama would work best for 4 Maravic creative writing. Additionally, it made Openai GPT-4O and Google Gemini 2.0 models better in tasks such as coding, Reasoning, Multilingual, Long-Contract and Image Generation. Meanwhile, Maverick struggled to live with the capabilities of the AI model, including Gemini 2.5 Pro, GPT -4.5 and Anthropic Cloud 3.7 sonnets.
While Meta claims that Bhamath can improve most of these models, except Gemini 2.5 Pro, the company is not able to reduce the hardware cost of training of its most powerful model.
Techcrunch said that Chinese AI company Deepsek has so far attracted attention from its competitor cheap model, which has put a meta on notice. The company is allegedly conducting intensive study of how the rival company developed its remarkable model, including the R1 and V3, including the former Lama at a lower operating cost than the model.
The company expanded that the Lama 4 scout model could run on an NVidia H100 GPU. The LLAMA 4 Maverick model can run on an Nvidia H100 DGX graphics system.
According to CNBC, Meta is ready to host its first Llamacon AI conference on 29 April. The company also has a standalone meta AI chatbot that is ready to launch in the second quarter of the year.
Meta is not the only ginger company with the timeline of its leading AI model. Openai recently adjusted the launch of its GPT-5 model, with the company’s CEO, Sam Altman, on social media that fans should expect new O3 and O4-MINI logic models as an alternative to GPT-5 in the coming weeks. The executive expanded that GPT-5 will now be launched in the coming months, which will give Openai extra time to bring the model to the standard.
