Yann LeCun, Meta's former chief AI scientist and a Turing Award winner, left the company at the end of 2025 after a 12-year partnership, marking a significant departure for one of artificial intelligence's most respected pioneers. The exit came after Meta's leadership restructured its AI division following a benchmarking controversy involving the Llama 4 model, and it signals growing internal tensions at a company trying to compete in the high-stakes AI race.

What Happened to Llama 4's Benchmark Scores?

Before Llama 4's launch, Meta's product team apparently manipulated benchmark results by using different versions of the model on different tests to inflate performance scores. When the issue became public, CEO Mark Zuckerberg responded by dismantling the entire division responsible for the benchmarking problems. However, LeCun ran FAIR (Facebook AI Research), Meta's long-term research laboratory, which was a separate unit from the product group that committed the benchmarking manipulation.

Despite having no involvement in the scandal, LeCun found himself caught in the fallout. Zuckerberg's restructuring placed him under the supervision of a 28-year-old former Scale AI CEO in a newly created "Superintelligence Labs" division. For a researcher with LeCun's stature and independent vision for AI development, this arrangement proved untenable.

How Did LeCun's Departure Unfold?

After leaving Meta, LeCun publicly confirmed what Zuckerberg had hoped to contain. He told the Financial Times that Llama 4's celebrated benchmark numbers were indeed "fudged a little bit," validating concerns about the model's credibility. This public acknowledgment compounded Meta's credibility problems in the AI community.

LeCun's departure has triggered broader consequences for Meta's AI ambitions. Major AI leaderboards now list "verification pending" on Meta's older models, and the company is experiencing what observers call a "Trust Tax." Developers and researchers who once viewed Meta as a serious AI research institution are reconsidering their involvement with the company.

What's Next for LeCun and Meta?

LeCun has already moved forward with his own venture, founding AMI Labs, which is generating significant attention in the AI research community. His new company is demonstrating "world models" that reportedly do not hallucinate, a major technical challenge that current large language models (LLMs) struggle with. An LLM is an artificial intelligence system trained on vast amounts of text data to understand and generate human language.

Meanwhile, Meta continues to grapple with the fallout from the benchmarking scandal and LeCun's high-profile departure. The company's open-source Llama model has been downloaded over 1.2 billion times, making it one of the most widely distributed AI models globally. However, the trust issues surrounding Meta's benchmarking practices threaten to undermine the credibility of future releases.

How to Understand the Broader Implications for AI Development

• Research Integrity: The benchmarking scandal highlights how pressure to compete in the AI race can lead teams to manipulate results, damaging trust in published performance claims across the industry.
• Organizational Structure: Placing a world-renowned researcher under a much younger executive with no connection to his work created friction that contributed to the departure of a key scientific leader.
• Open-Source Strategy: Meta's decision to open-source Llama was intended to democratize AI access, but the benchmarking controversy has raised questions about the reliability of the company's technical claims and future model releases.

The LeCun situation also reflects broader tensions in how AI companies balance rapid product development with scientific rigor. When Zuckerberg discovered the benchmarking problems, he responded with organizational upheaval rather than careful investigation of what went wrong and why. This approach cost Meta one of its most credible voices in AI research.

LeCun's departure is particularly significant because he was hired by Zuckerberg back in 2013 to establish FAIR, Meta's research division. For a decade, FAIR published significant academic papers that advanced the field, even though the lab never released consumer products. When ChatGPT launched publicly in late 2022, Zuckerberg realized Meta had fallen behind in the AI race and pivoted to an aggressive open-source strategy with Llama.

"Yeah, the celebrated Llama 4 numbers were fudged a little bit," LeCun told the Financial Times after leaving Meta.

Yann LeCun, Former Chief AI Scientist at Meta

The irony is that LeCun's departure may ultimately benefit him. His new company, AMI Labs, is already generating significant hype in the AI research community, and he is no longer constrained by Meta's organizational politics or the company's credibility challenges. For Meta, losing a Turing Award winner and one of the founding fathers of modern AI represents a significant blow to its scientific credibility at a critical moment in the AI industry.