WSJ: Amid struggles, Meta delays release of Llama 4 “Behemoth” AI model
The jumbo version of Meta’s flagship AI model Llama has been delayed amid performance struggles.
According to the report from The Wall Street Journal, Meta engineers aren’t able to achieve the kinds of exponential leaps in performance that early generations of generative-AI models delivered, a phenomenon that is manifesting across the industry.
Recent breakthroughs in efficiency and cost achieved by breaking models down into smaller parts — like DeepSeek — are matching or beating scores of some of the drastically larger and more expensive models, like Meta’s Llama and OpenAI’s o1.
This shift has occurred as some tech giants like Microsoft appear to be pulling back on large commitments for AI data centers.
Meta recently upped its capex projections to $60 billion to $72 billion for this year, including a massive city-sized data center in Louisiana. The company’s gargantuan data center plans are built on the expectation of running the Llama 4 models.
Meta released two smaller versions of Llama 4 in April: “Maverick” and “Scout.”
The report says the delay is causing frustration within the company, as it has been hyping up expectations for the new model over the past year:
“Senior executives at the company are frustrated at the performance of the team that built the Llama 4 models and blame them for the failure to make progress on Behemoth, according to people familiar with their views. Meta is contemplating significant management changes to its AI product group as a result, the people said.”
Meta shares were down 2.4% on the news.
Recent breakthroughs in efficiency and cost achieved by breaking models down into smaller parts — like DeepSeek — are matching or beating scores of some of the drastically larger and more expensive models, like Meta’s Llama and OpenAI’s o1.
This shift has occurred as some tech giants like Microsoft appear to be pulling back on large commitments for AI data centers.
Meta recently upped its capex projections to $60 billion to $72 billion for this year, including a massive city-sized data center in Louisiana. The company’s gargantuan data center plans are built on the expectation of running the Llama 4 models.
Meta released two smaller versions of Llama 4 in April: “Maverick” and “Scout.”
The report says the delay is causing frustration within the company, as it has been hyping up expectations for the new model over the past year:
“Senior executives at the company are frustrated at the performance of the team that built the Llama 4 models and blame them for the failure to make progress on Behemoth, according to people familiar with their views. Meta is contemplating significant management changes to its AI product group as a result, the people said.”
Meta shares were down 2.4% on the news.
