Meta Creates New Avacado AI Model, Know Details  - Inspirepreneur Magazine

Meta Creates New Avacado AI Model, Know Details 

Shivangi
Dec 10, 2025 4:17 PM IST
Category America
Meta Creates New Avacado AI Model, Know Details 

Synopsis

Meta is working on an AI system called Avocado slated for release in the first quarter of 2026 marking a change from Zuckerberg's pledge to support open-source models. This change follows the feedback on Llama 4 and disagreements over licensing. Meta has put $14.3 billion into Scale AI. Named its founder, Alexandr Wang as chief AI officer. Samsung ceased utilising Llama 4 after Meta requested caution. Google unveiled Gemini 3 OpenAI introduced GPT-5.1 updates and Anthropic revealed Claude Opus 4.5. Meta raised its 2025 capital expenditure forecast to $70-72 billion. The firm cut 600 jobs with employees reporting 70-hour workweeks. Meta declined to verify Avocado's nature.

Meta is creating a custom AI model called Avocado. The​‍​‌‍​‍‌​‍​‌‍​‍‌ company plans to unveil the prototype in 2026. This is a departure from CEO Mark Zuckerberg's usual practice of strictly employing open-source AI. The switch happens as the firm is dealing with a lot of uncertainty, difficulties in obtaining licenses and increasing competition from OpenAI, Google and ​‍​‌‍​‍‌​‍​‌‍​‍‌Anthropic.

01
Chapter one

A Big Change in Strategy

This development represents a change from Meta’s position as a proponent of open access to AI. Early in 2025, Zuckerberg emphasised the company’s Llama series of open-source models. Yet in Meta’s October earnings call, he referenced Llama a single time.

This shift in approach follows the response to Llama 4, which did not generate enthusiasm among developers and raised doubts about Meta's open-source approach. In June 2025, Meta spent $14.3 billion to acquire a stake in Scale AI. The 28-year-old founder, Alexandr Wang, was appointed as AI officer. Wang currently leads Meta Superintelligence Labs together with ex-GitHub CEO Nat Friedman, who manages product development and practical research.

02
Chapter two

Licensing Issues with Partners

The licensing conditions have sparked friction with collaborators. Firms catering to over 700 million active users require special permission from Meta to utilise Llama 4. In November, Meta advised Samsung's semiconductor division, which had been employing Llama 4, to proceed carefully. Subsequently, the Korean electronics giant ceased its use of the model and started enhancing its proprietary Gauss system. A Samsung representative told journalists that the company wished to avoid needless conflict with Meta.

The urgency has grown as rivals launch products. Google rolled out Gemini 3 in November. OpenAI revealed upgrades. Anthropic launched Claude Opus 4.5. Meta raised its 2025 capital expenditure forecast to $70-72 billion. It anticipates greater investment in 2026.

03
Chapter three

Staff Cuts and Long Hours

This comes after Meta cut 600 positions at Meta Superintelligence Labs in October as part of efforts to simplify operations. Teams are experiencing 70-hour workweeks. "I think we’ve already created the lab with the concentration of talent in the sector," Zuckerberg mentioned during the October earnings report.

Meta refused to provide details on whether Avocado will be private. A representative merely stated that the training of the model is progressing as expected. The absence of communication has added to the internal uncertainty. Employees are still unclear about the company's direction regarding AI.

04
Chapter four

Why Is This Important 

For a time, he promoted open-source AI as the way forward. He argued that providing AI models freely to all would speed up innovation and help everyone. Numerous researchers and developers created projects based on Llama models. Currently, Meta seems to be heading in the right direction.

Moving to a framework indicates Meta's strong desire to retain the top AI advancements exclusively for itself. This change would, among other things, probably provide the company with greater authority over the technology and more opportunities to profit from it. However, this stance contradicts Zuckerberg's statements about open-source AI being more beneficial for the world.

Rivals such as OpenAI and Anthropic have consistently maintained their models as confidential. Google utilises a combination of proprietary models. Should Meta choose to make Avocado entirely private, it would be aligning itself with the approach taken by the majority of its competitors.


More news about technology companies and updates on AI development can be found in Inspirepreneur Magazine. Stay informed about major shifts in artificial intelligence strategy.

Written by Shivangi

At Inspirepreneurs Magazine, covering entrepreneurship, business failures, and the human stories behind the world's most ambitious founders. She writes at the intersection of strategy and storytelling.