Meta delays rollout of flagship AI model amid internal performance concerns

The WSJ is reporting that Meta has delayed the release of its flagship large language model, Llama 4 “Behemoth,” citing internal concerns about its performance and whether it offers meaningful improvements over previous versions.

• Delayed Release: Behemoth, originally targeted for an April release, has now been pushed to fall or later.

• Performance Issues: Engineers say Behemoth shows minimal gains over prior models, sparking debate about whether it’s ready for public release.

• Internal Frustration: Meta executives are frustrated with the Llama 4 team’s progress and may restructure the AI product group.

• Mismatch with Public Claims: Behemoth has been publicly touted as outperforming rivals, but insiders say training issues are limiting its real-world performance.

• Earlier Releases Questioned: Two Llama models released in April performed well on AI leaderboards, but Meta later admitted it had optimized those submissions specifically for the tests.

• Research Team Turnover: 11 of the 14 researchers behind the original Llama model have since left the company.

• Industry Trend: Meta’s delays reflect broader challenges in the AI space—OpenAI’s GPT-5 and Anthropic’s Claude 3.5 Opus have also missed initial release expectations.

• Meta’s AI Spending: The company plans up to $72 billion in capital expenditures this year, much of it devoted to AI development.

Shares of Meta are trading down -$14.12 or -2.16% 15 $645.15, after trading as high as +$3.89 on the day.

This article was written by Greg Michalowski at www.forexlive.com.