Outgoing Meta Chief AI Scientist Yann LeCun admitted in a Financial Times interview that benchmark results for the Llama 4 model were manipulated. LeCun, who departs Meta after a decade, confirmed the team used different models across various tests to artificially boost scores before the model's April 2025 release.
The revelation caused significant internal strife. CEO Mark Zuckerberg reportedly expressed anger, leading to the marginalization of the entire GenAI team. Many members of the team have since departed the company.
The Llama 4 controversy and fundamental disagreements with Zuckerberg over the company’s AI strategy are cited as primary reasons for LeCun’s departure. This event casts doubt on Meta’s future AI development and has created a major trust crisis for the company.