According to a report by The New York Times, Meta's next-generation foundational AI model, codenamed "Avocado," has postponed its planned release from this month (March 2026) to at least May due to underwhelming performance in internal testing. While the model outperforms Meta's previous models and Google's Gemini 2.5 in reasoning, coding, and writing capabilities, it still lags behind the Gemini 3.0 released last November and cannot compete with the top models from Google, OpenAI, and Anthropic. As a result, Meta AI leadership is discussing the possibility of temporarily licensing $Google Gemini to power its own AI products (such as the Meta AI chatbot), but no decision has been made yet.