Meta is once again intensifying its push into artificial intelligence, and this time the company is working on two next-generation AI models codenamed Mango and Avocado. The social media giant is reportedly developing these advanced systems to challenge the current dominance of OpenAI and Google in the AI space.
If development stays on track, both models are expected to debut in the first half of 2026, marking a major milestone in Meta’s long-term AI strategy.
According to The Wall Street Journal, Meta’s new projects represent a significant new chapter in the company’s AI ambitions.
Mango is focused on image and video generation
Avocado is being built as a text and coding-focused large language model (LLM)
Together, these models aim to cover creative generation, reasoning, and advanced technical problem-solving—areas where competition has intensified rapidly.
The initiative is being led by Meta’s Chief AI Officer Alexandr Wang, the 28-year-old founder of Scale AI, who joined Meta earlier this year after the company acquired a near-majority stake in his startup for over $14 billion.
Mango and Avocado will be the first major outputs from Meta Superintelligence Labs (MSL), a newly formed division created to accelerate Meta’s AI breakthroughs. The lab is designed to operate with greater speed, autonomy, and focus than Meta’s previous AI units.
This investment reflects CEO Mark Zuckerberg’s determination to claw background from rivals who have dominated the AI conversation over the past year.
Mango is being developed as a high-quality image and video generation model, with the goal of competing directly with:
OpenAI’s Sora
Google’s Gemini 3 Flash
The model is expected to deliver advanced creative outputs tailored for creators, developers, and businesses, positioning Meta strongly in the rapidly growing generative media space.
Avocado is being designed as Meta’s next major large language model, with a strong focus on:
Advanced coding capabilities
Enhanced reasoning and logic
Technical problem-solving
During an internal Q&A, Wang described Avocado as Meta’s most ambitious LLM to date. He reportedly said:
“We’re building models that understand the world, not just words,”
highlighting Meta’s early work on ‘world models’.
World models represent a shift away from traditional AI systems that rely mainly on predicting the next word in a sentence. Instead, they aim to give AI systems a more perception-based understanding of reality, which experts widely consider the next frontier in artificial intelligence development.
Meta’s aggressive AI push follows a major internal reorganisation earlier in 2025, when the company consolidated its AI divisions and placed Wang in charge of a newly formed elite research group.
As part of this effort:
Meta has poached more than 20 top scientists from OpenAI
The company has assembled a “super team” of over fifty AI specialists under the Meta Superintelligence Labs banner
This restructuring gives Meta the talent depth needed to compete with the industry’s leading AI labs.
The timing of Meta’s push is critical, as the AI arms race is accelerating rapidly.
Google recently released Gemini 3 Flash, a faster and more affordable version of Gemini 3 Pro aimed at mass adoption
OpenAI continues to expand its ecosystem with ChatGPT Images and Sora, its video-generation platform that has gained rapid popularity among creators
Earlier this year, Meta launched Vibes, a video generator developed in collaboration with Midjourney. However, OpenAI responded within days by rolling out Sora, underlining how competitive the image and video generation space has become.
Google’s momentum has only added to the pressure. The Gemini suite reportedly grew from 450 million monthly users in July to over 650 million by October, largely driven by the rollout of Gemini Nano and subsequent models.
Inside OpenAI, CEO Sam Altman reportedly declared a “code red” after Google unveiled Gemini 3, accelerating OpenAI’s own development timelines.
With Mango and Avocado now in development, Meta is preparing to re-enter the AI spotlight in a serious way. The company’s heavy spending, combined with Wang’s leadership, aggressive hiring, and structural overhaul, signals that Meta is no longer content to trail behind competitors.
Instead, Meta aims to stand toe-to-toe with Google’s Gemini models and OpenAI’s GPT lineup, armed with a powerful mix of creative AI tools and deep computational intelligence.