Meta delays release of new AI, weighs licensing Google's Gemini after disappointing trial runs: report

Core Insights - Meta is delaying the release of its new AI model, "Avocado," by approximately two months due to underperformance in internal tests compared to competitors like Google, OpenAI, and Anthropic [1][2][3] - The company is considering temporarily licensing Google's Gemini to enhance its AI products following the disappointing results [1][3] Development and Investment - Meta has invested significantly in AI, spending billions on hiring top talent and committing $600 billion to build data centers, with a projected expenditure of up to $135 billion in 2023, nearly double the $72 billion spent in 2022 [3] - The development of Avocado has been ongoing for months, with the model expected to launch around May instead of March due to its inability to meet performance benchmarks in reasoning, coding, and writing [3][10] Performance Comparison - Avocado has shown improved performance over Meta's previous AI model but still lags behind Google's Gemini 3.0, which was released in November [4] - Despite setbacks, Meta's leadership remains optimistic about the trajectory of their AI models, with expectations for significant advancements in the near future [4][7] Organizational Changes - Meta's AI division, led by Alexandr Wang, has faced some turnover, with researchers leaving before the release of Avocado [12] - The company is restructuring its AI teams, creating a new team under Andrew Bosworth to collaborate with Wang, indicating a strategic shift in its AI development approach [15][16] Competitive Landscape - Meta's AI efforts are closely monitored within the competitive landscape of AI development, where Google, OpenAI, and Anthropic are perceived as leaders [7] - The company has historically supported open-source models, which contrasts with the more restrictive approaches of its competitors [15]