Meta's 'Avocado' AI Model Delayed Amid Performance Issues Against Competitors

META has delayed the launch of its next-generation AI model, codenamed "Avocado," from March to at least May 2026 after internal testing revealed it underperforms key competitors. The model reportedly falls between Google's Gemini 2.5 and Gemini 3.0 in benchmarks, lagging behind the latest models from Google, OpenAI, and Anthropic on crucial reasoning, coding, and writing tasks. Meta's share price declined on the news as investors weighed the implications for the company's AI strategy.

The delay carries broader strategic significance. Under CEO Mark Zuckerberg's mandate to pursue artificial general intelligence, Meta appears to be shifting away from its long-standing open-source approach with the Llama model family toward proprietary development. Executives have reportedly discussed the possibility of temporarily licensing technology from Google's Gemini models to support some of Meta's AI products while Avocado is refined, though no final decision has been made.

For investors, the setback raises questions about Meta's ability to compete at the frontier of AI development despite its massive capital expenditure plans. The potential pivot from open-source to proprietary models could alter Meta's competitive positioning and developer ecosystem. Markets will be watching closely for any partnership announcements with GOOGL and whether the delayed May timeline holds.