Meta plans Mango and Avocado AI models for 2026 launch

Meta is preparing to introduce two new artificial intelligence models, Mango and Avocado, as part of a renewed push to challenge industry leaders OpenAI and Google, according to a report by The Wall Street Journal. The models are expected to launch in the first half of 2026 and represent a significant escalation of the company’s long-term AI strategy.

The report stated that Mango is being developed as an image and video generation model, while Avocado is positioned as a text and coding-focused large language model. Both projects are being led by Meta’s Chief AI Officer Alexandr Wang, the founder of Scale AI, who joined the company earlier this year after Meta acquired a near-majority stake in his startup for more than $14 billion.

Mango and Avocado will be the first major outputs from Meta Superintelligence Labs, a newly formed division created to accelerate the company’s AI development efforts. The move signals Chief Executive Officer Mark Zuckerberg’s intent to regain ground after rivals have dominated the AI landscape over the past year.

As per the report, Mango is designed to deliver high-quality image and video generation and is expected to compete directly with OpenAI’s Sora and Google’s Gemini 3 Flash. Avocado, meanwhile, is being built to improve Meta’s capabilities in reasoning and coding, areas where the company’s existing Llama models are considered less competitive.

The push follows a major internal reorganisation earlier in 2025, when Meta consolidated its AI teams and placed Wang in charge of an elite group of researchers and engineers. The company has also recruited more than 20 senior scientists from OpenAI, assembling a team of over 50 specialists under the Meta Superintelligence Labs banner, the report informed.

Earlier this year, Meta launched Vibes, a video generation tool developed in collaboration with Midjourney. However, OpenAI’s subsequent release of Sora highlighted the intensifying competition in AI-generated video and creative tools, underscoring the urgency behind Meta’s next wave of model development.