Meta’s AI chief Alexandr Wang faces make-or-break moment with Meta’s next AI launch

Wang

Meta is having a rough time behind the scenes, and the pressure inside the company is starting to show.

At the heart of the drama is Meta’s big bet on artificial intelligence. Mark Zuckerberg has poured huge money and resources into building an elite AI team, but that team is now under intense scrutiny after the company’s recent missteps. The spotlight is firmly on Meta’s next AI model, known internally as Avocado, which is meant to replace the Llama series and help Meta catch up with rivals.

Things became tense after the launch of Llama 4 earlier this year. Inside the company, many felt the rollout did not go as planned. The model was criticised for weak real-world performance and poor coding skills, and there were also allegations that some of its test results were dressed up to look better than they actually were. While Meta did not publicly dwell on the criticism, employees say it triggered a wave of concern at the top.

Much of that pressure has landed on Alexandr Wang, Meta’s chief AI officer. At just 28, Wang leads a small but powerful group called TBD Lab, where Avocado is being built. According to people familiar with the situation, Wang is now under constant pressure to deliver a model that can stand up to Google, OpenAI and Anthropic, all of which have released well-received AI products in recent months.

Avocado itself has reportedly hit several roadblocks. The model is still struggling with important performance checks, and as a result, its launch has been delayed. It was earlier expected in 2025, but is now likely to arrive in early 2026. For a company trying to prove it can move fast in AI, that delay has not gone down well internally.

Wang is not alone in feeling the heat. Nat Friedman, the former GitHub chief who now holds a senior role at Meta AI, is also under pressure to deliver something people actually want to use. Under his watch, Meta launched Vibes, a stream of AI-made short videos. Employees say it was rushed out and ended up looking far less impressive than similar tools from OpenAI, especially when it came to realistic voices and facial movement.

The strain is being felt across the company. Employees say 70-hour work weeks have become common, even as Meta cut hundreds of jobs and reorganised teams throughout the year. There is also growing disagreement inside Meta about how things should be built. Some leaders want to move fast and experiment more, while others worry that speed is coming at the cost of quality.

For now, Avocado has become a make-or-break moment for Meta’s AI ambitions. If it succeeds, it could quiet the criticism. If it fails, the questions around leadership, direction and strategy are only going to get louder.