Meta is redirecting its AI strategy with the development of a new model dubbed Avocado, as per Bloomberg reports. This shift involves moving towards a closed, paid system with tighter controls on access. The Avocado model is expected to launch in early 2026, marking a departure from Meta’s previous open-source approach.
The decision to pivot towards a paid model comes amidst increasing development costs and heightened competition in the AI landscape. Meta’s CEO, Mark Zuckerberg, hinted at this change earlier in the year, emphasizing the need to be cautious about open-sourcing technologies for safety reasons.
The Avocado model represents a significant internal reset for Meta, following the discontinuation of the Llama 4 project, internally known as Behemoth. To drive this new direction, Meta appointed Alexandr Wang as the Chief AI Officer, who leads the TBD Lab responsible for developing the Avocado model. Wang’s expertise in closed-model strategies aligns with Meta’s new direction towards monetizing its AI products.
Zuckerberg has instructed teams to utilize rival systems like Google’s Gemma and OpenAI’s GPT-OSS in the training pipeline for the Avocado model. This collaborative approach aims to enhance development processes and ensure commercial competitiveness.
The anticipated launch of the paid Meta AI model was initially expected before the end of 2025 but has now been rescheduled to the first quarter of 2026. This strategic shift towards a paid model signifies Meta’s commitment to driving innovation and monetizing its AI advancements.