Meta Postpones Avocado AI Model Launch Following Internal Testing Results Revealing Performance Gaps with Google and OpenAI

Meta has postponed the release of its AI model, “Avocado,” initially scheduled for mid-March 2026, now expected by May at the earliest. Internal tests revealed that Avocado lagged behind competitors like Google, OpenAI, and Anthropic in performance metrics, though it outperformed Meta’s prior models and Google’s Gemini 2.5. There’s been discussion about temporarily licensing Gemini, but no concrete decision has been reached. Additionally, a next-generation model named “Watermelon” is already in the pipeline, along with an image and video generator dubbed “Mango.” Meta plans to release updates soon, despite losing momentum with its Llama 4 model after initial success. CEO Mark Zuckerberg has invested heavily in AI, totaling $14.3 billion in Scale AI, with CEO Alexandr Wang leading Meta’s frontier AI division, “TBD Lab.” There are indications that Meta may shift away from its open-source approach. Stay tuned for future updates on Meta’s AI advancements.

Source link