Meta released the first two models of the Llama 4 Middle Employer: Llama 4 Scout and Llama 4 MAVERICK. The company said in “Mavrick” is the “spine” from Monday and outperforms the understanding of the images and the text “for public use and chat situations.” Blog postAlthough the smaller model of models can address things like “multi -sensitive summary, uniformity of a wide -range user activity for personal tasks, thinking about the vast code bases.” The company also presented Llama 4 Beheemoth, a coming model that says it is “among the smartest LLMS in the world” – and CEO Mark Zuckerberg He said that we will hear about a fourth model, Llama 4 logic, “next month.”
Both MAVERICK and Scout are available for download now from Lama The site and EmbroideryIt has been added to Meta AI, including WhatsApp, Messenger and Instagram DMS.
Meta says that Scout has 17 billion active teachers with 16 experts. According to Zuckerberg, “It is very fast, multimedia, and has a leading industry in an almost countless symbolic context, and is designed to run on a single graphics processing unit.” On the other hand, it has 17 billion active teachers with 128 experts. The company says it exceeds my competitors such as GPT-4O and Gemini 2.0 on coding, thinking and multi-language image standards and long context, and is stacked against Deepseek v3.1 on thinking and coding.
Zuckerberg is already called the next giant model, which is still trained, “the highest basic performance model in the world”, with 288 billion active teachers, according to the company. This may not be here yet, but we will likely hear a lot about that and the thinking model soon; Meta Grand IQ conference, For the cousinsIt is just a few weeks away.
This article originally appeared on Engadget on https://www.engadget.com/Ai/TETA-INTRODUSES-LAMA-4-With-two-New-Models-aviliable-l-nd-two-more-on-The-adway-21524295.html?
[og_img]
Source link