Meta is preparing a major comeback to the artificial intelligence industry with two new models named Mango and Avocado, expected to launch in the first half of 2026. These models are designed to challenge the dominance of OpenAI and Google, who currently lead the AI landscape. Mango will focus on generating high-quality images and videos, while Avocado will serve as Meta’s next large language model, built for text generation, complex reasoning, and coding capabilities. Both models are being developed as part of the company’s new Meta Superintelligence Labs, led by Chief AI Officer Alexandr Wang.

Wang was the founder of Scale AI, and his move to Meta has already influenced a major shift within the company. Meta has reorganised internal divisions and recruited over 20 leading scientists from OpenAI to accelerate progress. Mango aims to compete with OpenAI’s Sora and Google’s Gemini 3 Flash in visual creation, while Avocado is expected to surpass the performance of Meta’s existing Llama models. Wang has described Avocado as a step toward world-model understanding, where AI grasps real-world context instead of simply predicting text patterns.

The timing of Meta’s announcement comes as competition between major AI players intensifies. Google’s Gemini models are growing rapidly, and OpenAI continues to expand with products like ChatGPT Images and Sora. Meta has already tested the waters with its Vibes video generator but now appears ready to take a more aggressive approach by investing in long-term research and development. Mango and Avocado represent Meta’s goal to reclaim leadership in AI and reshape the market with next-generation creative and reasoning capabilities. If successful, 2026 could mark a turning point in the company’s AI journey and significantly shift the future direction of the industry.