2026 Meta Mango And Avocado AI Launch: Meta Platforms is gearing up for an epic showdown with AI powerhouses OpenAI and Google in 2026. The social media giant, however, is working on introducing a new generation of AI models that would outpace its competitors as the company unveiled in a report from The Wall Street Journal on Thursday.
Meta executives unveiled a roadmap in an internal Q&A on Thursday that outlined two major projects scheduled for launch in the first six months of 2026. This forward planning should signal an end to the near-identical updates of Llama’s current family of models.
Mango and Avocado: A Two-Pronged Attack
The new effort is being led by two separate models with code names that suggest a more open-minded approach.
- “Mango”: This is a high quality (non-cat) image/video focused AI model. It reflects Meta’s desire to compete head on in the rapidly shifting generative media space, which has been dominated by tools from OpenAI and Google.
- “Avocado”: This will be Meta’s next big text-based large language model (LLM). Meet Avocado: Meta’s plan to crush rivals’ AI dreams In a blog post, Chief AI Officer Alexandr Wang said Avocado will be heavily focused on advanced coding capabilities, addressing a traditional weak spot Mandarin learners of previous Meta models as compared with its competitors.
Also Read: Google Just Dropped Gemini 3 Flash: Frontier Intelligence at Warp Speed
The New “Superintelligence Labs”
The models are the first major product of Meta Superintelligence Labs (MSL), a dedicated division formed after a wide-reaching, company reorganization earlier this summer.
MSL’s formation highlights CEO Mark Zuckerberg’s aggressive push to eliminate the AI deficit. For this new division Zuckerberg went on a recruitment binge, hand-picking researchers himself and hiring more than two dozen of them from OpenAI.
Chief AI Officer as Head of the Whole Effort Most notably, he hire 28-year-old Scale AI founder Alexandr Wang to head up and orchastrate the entire effort as Chiefs AI Officer.
This strategic shift was reinforced by Meta’s widely reported $14 billion+ commitment to nearly control Scale AI, with Wang now seen as the main driver of Meta’s post-Llama AI strategy.
Beyond LLMs: The Push for “World Models”
In a sign of even higher ambition, Wang also said on the internal session that Meta is in the early stages of working on “world models,” according to CNBC.
Whereas existing language models such as LLMs predict the next word in a sequence given text, world models are trained to reason about and simulate physical laws using large-scale visual observation. This is orders of magnitude closer to ‘commonsense AI’.
Meta’s stock was largely flat in after-hour trading after the report on Tuesday, though shares were up 2.3% during the day which might suggests that investors see a silver lining for the company long-term as an AI business.
Also Read: Wan 2.6 Dropped: Scripting To Final Video, It Does It All in One Take
