Meta Platforms Inc. is accelerating its artificial intelligence ambitions with two new models—Mango for images and videos, and Avocado as its next-generation large language model—both slated for release in the first half of 2026. The initiative, detailed in an internal meeting led by Chief AI Officer Alexandr Wang, marks a departure from the company’s open-source Llama series amid intensifying competition from OpenAI and Google. According to the Wall Street Journal, Wang outlined the roadmap during a companywide Q&A, signaling Meta’s push into proprietary technology to regain ground in the AI race.

Wang, who joined Meta earlier this year after selling his Scale AI stake in a $14 billion deal, emphasized the models’ potential during the session. ‘Mango will be our high-stakes image and video model,’ he said, per sources familiar with the discussion cited by the Journal. Avocado, meanwhile, represents Meta’s bid for a frontier text-based LLM to challenge leaders like OpenAI’s GPT series. This dual-track development comes as Meta grapples with internal disruptions from its strategic shift, reported widely in recent weeks.

The announcement has stirred debate among engineers and executives, with some viewing it as a bold corrective to Llama 4’s underwhelming performance. CNBC, in a December 9 article, highlighted ‘internal friction’ over the Avocado push under Wang’s leadership, noting tensions as Meta races rivals while reallocating resources from open-source efforts.

Wang’s Rise and Meta’s $14 Billion Bet

Alexandr Wang, once the world’s youngest self-made billionaire at 24, brings a data-labeling pedigree from Scale AI to Meta’s Superintelligence Labs. His Wikipedia profile details his roots in Los Alamos, New Mexico, as the son of physicists, and his early prowess in math olympiads. Meta’s acquisition of Scale talent, valued at $14 billion, aimed to bolster data infrastructure for training massive models, but Business Insider reported fallout including pay cuts and poaching by rivals just five months in.

Wang’s recent X posts underscore his focus on generative media. In September 2025, he teased ‘Vibes,’ a Meta AI app feature for AI-generated short-form videos, amassing millions of views. Such innovations hint at Mango’s underpinnings, potentially leveraging Scale’s expertise in curating multimodal datasets.

Investing.com echoed the Journal’s reporting on December 18, confirming Meta’s work on Mango as a video-centric model alongside Avocado, citing internal sources. This aligns with Digitimes’ insight that Avocado signals a ‘shift to closed development,’ with a spring 2026 target.

From Llama Stumbles to Proprietary Push

Meta’s pivot stems from Llama 4’s internal ‘failure,’ per WinBuzzer’s December 9 analysis, prompting abandonment of open-source roots for closed models like Avocado. The Los Angeles Times detailed this high-stakes shift toward monetizable AI, contrasting Zuckerberg’s vision with execution challenges.

Moneycontrol on December 13 framed Avocado as Wang’s ‘make-or-break’ moment, amid setbacks like model delays and leadership critiques. TechTimes reported Wang questioning Zuckerberg’s hands-on style, fueling upheaval as Meta invests billions in compute and talent.

ZDNet noted Nvidia’s opportunistic move into transparent open models like Nemotron 3, filling a void as Meta retreats. Wang’s strategy bets on superior proprietary performance to justify the secrecy.

Internal Turbulence and Strategic Risks

CNBC detailed confusion from the Avocado transition, with teams split between legacy Llama work and new closed efforts. Financial Express questioned if Meta’s $15 billion Wang bet will pay off, tying Avocado’s success to catching AI leaders.

Scale AI’s post-deal woes, including client losses covered by Business Insider, ripple into Meta, where rivals lure talent with better terms. Wang’s X activity, like promoting SAM Audio releases, shows ongoing momentum despite headwinds.

The Journal’s sources indicate Mango could integrate with Meta’s social platforms for enhanced content generation, potentially unlocking ad revenue. Avocado aims for parity with top LLMs in reasoning and multimodality.

Timeline and Competitive Pressures

Both models target H1 2026, per WSJ and Investing.com, aligning with Meta’s aggressive cadence. Digitimes specified spring for Avocado, positioning it against anticipated OpenAI and Google releases.

Wang’s leadership, once hailed, now faces scrutiny. Moneycontrol cited growing pressure post-setbacks, while TechTimes highlighted critiques of Zuckerberg’s involvement. Meta’s AI spend, exceeding peers, underscores the wager.

On X, sentiment mixes excitement with skepticism; Wang’s posts on AI media generation draw engagement, but industry chatter questions closed-model viability against open alternatives like Nvidia’s.

Implications for AI’s Closed Frontier

Meta’s move could reshape industry norms, prioritizing control over community-driven progress. As WinBuzzer noted, ditching Llama openness invites copycats but promises tailored enterprise tools.

LA Times emphasized monetization focus, with Mango enabling creator economies via video AI. Success hinges on Wang delivering benchmarks surpassing GPT-5 equivalents.

For insiders, this pivot tests whether proprietary walls yield defensible moats in an era of rapid iteration, with Meta’s fortunes tied to these fruit-named flagships.