By Breaking Editor 
Meta Platforms Inc., the world’s largest social media network and led by Jewish American, Mark Zuckerberg, has postponed the launch of its highly anticipated artificial intelligence model, “Avocado,” to at least May 2026, according to sources reported by the New York Times.
Originally scheduled for a March release, the delay comes after internal tests indicated that the model does not yet meet the performance benchmarks set by competitors such as Google, OpenAI, and Anthropic in areas including reasoning, coding, and writing.
Meta’s current focus on artificial intelligence is a life and death strategy marked by significant investments. If it fails to evolve, the technological imagination of the public may abandon their gifs and surveillance advertising platforms.
At the center of their efforts is the highly anticipated AI text-generation model, “Avocado,” which was originally scheduled for release in March 2026 but has now been delayed until at least May. The delay stems from internal evaluations revealing that Avocado underperforms compared to the Gemini 2.5 and 3.0 models developed by Google. As a temporary solution to address this capability gap, Meta is even considering licensing Google’s Gemini model.
Meta is likely optimizing the media advertising flank of its operations. For this, it needs an LLM layer capable of automatically generating good copy that consumers will pay for due to convenience.
In this high interest rate environment, Meta’s broader AI agenda is underpinned by massive capital expenditure plans, with a budget of $115–$135 billion for 2026, a significant jump from the $72 billion spent the previous year. These investments include building advanced data centers like “Prometheus,” a 1GW facility scheduled to go live in 2026, and “Hyperion,” a planned 5GW center expected to come online by 2030. In addition to infrastructure, Meta has entered into multi-billion-dollar partnerships with leading computing firms such as Nvidia, AMD, and Coreweave to ensure sufficient compute capability for its AI endeavors.
The company’s development strategy extends to acquiring companies like Manus for $2.5 billion, along with Moltbook and Scale, to enhance its AI and robotics capabilities. Reports also indicate that Meta allocates approximately $1 billion annually for salaries to support its elite AI research teams, such as the specialists at TBD Labs.
Apart from Avocado, Meta is reportedly developing other AI models, including “Mango,” which focuses on image and video generation, and “Watermelon,” a multimodal system designed to process multiple data types concurrently. These models add to an increasingly diversified AI portfolio, reflecting Meta’s commitment to pushing the boundaries of technology.
Additionally, Meta plans to integrate AI into its products and platforms. One area of focus is robotics, with job postings suggesting work on humanoid robotics projects. Another area involves wearables like Ray-Ban smart glasses, which could capture egocentric and 3D data to support robotics training and enhance intelligence for AI applications. Furthermore, Meta intends to embed AI across its suite of apps under the “Family of Apps” initiative, similar to the integration strategies employed by Google.
Despite its ambitious vision, Meta faces criticism and challenges. Observers criticize certain leadership decisions, such as the appointment of Alexander Wang to lead the Avocado project, and accuse Meta of failing to innovate independently. Competitors like ByteDance are seen as more advanced in enterprise AI, and there is skepticism about whether Meta’s enormous investments in data acquisition and stateless models can deliver transformative breakthroughs.