Meta postpones the launch of Llama 4 Behemoth: the model is still not ready
Meta is once again postponing the release of its flagship AI model Llama 4, codenamed Behemoth.
According to The Wall Street Journal, the company planned to present it back in April at the LlamaCon conference, but the launch was first postponed to June and now to autumn or even later.
Here's What We Know
The reason for the delay is difficulties with model improvement. Meta engineers are allegedly still "struggling" to improve its capabilities, despite Mark Zuckerberg's loud statements that Behemoth is "the most powerful basic model in the world". Questions are already arising within the company as to whether the updates are really significant enough to justify a public release.
So far, Meta has released two lighter variants of the Llama 4 - Scout and Maverick - and announced another compact model with the internal name Little Llama. Behemoth, according to the company, has 288 billion active parameters and outperforms GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on STEM benchmarks.
Despite the promises, Meta has not provided any exact release dates for the model. During the LlamaCon conference, Behemoth was mentioned only in passing, with no news about the launch date. This is worrisome, especially considering Meta's large-scale investments in AI - the company plans to spend up to $72 billion on AI infrastructure this year.
Recently, Meta officially launched the first standalone Meta AI app - users can now interact with the artificial intelligence not only on Facebook, Instagram, WhatsApp, and Messenger, but also through a dedicated app.
Source: The Wall Street Journal
