Meta’s ‘Behemoth’ Lama 4 Model may be still months away
Last month, meta Its first productive AI meeting. Although the program offered some important improvements to developers, the company considered the AI to consider how important it was. Now, we know a little about why, why, why, thank you In The Wall Street Journal.
According to the report, Meta initially planned to launch his “Behemoth” Lama 4 model in the April developer event, but later delayed its release in June. Now, it has been pushed back again until “fall or later”. Meta Engineers Mark Zuckerberg struggles to significantly improve the model’s capabilities that “the world’s most performed base model”.
Meta has already released two small Lama 4 models, And the fourth lighter model with the nickname “Little Lama” is also teased. Meanwhile, the “Behemoth” model consists of 288 billion active parameters and “GPT -4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro will be overcome on several stems benchmarks” .
Meta has never given a tough timeline when to expect the model. The company said it was “still training”. While Bheemoth got some nodes during the Lamacon keynote, there were no updates on when it could be ready. Perhaps it could be several months. There are clear questions inside the meta “whether improvements on pre -versions are important to justify public release.”
Meta did not respond immediately to the request for the comment. As the report says, this is not the first company to enter snags as it is in the race to release new models and overcome competitors. When it comes to AI, the delay is still noticeable due to the superior ambitions of the meta. Zuckerberg did AI To spend more with meta planning This year on its AI infrastructure.
