
Nine months and $14.3 billion after hiring Alexandr Wang and forming an AI super team, Meta's AI masterplan is looking increasingly shaky.
The company's latest AI model, code named Avocado, is apparently being delayed until May, instead of launching this month, according to a
report in the New York Times on Thursday. While Avocado performed better than Meta's previous AI model, and better than Google Gemini 2.5, its performance lags the newer version of Gemini, 3.0, which came out in November, as well as the latest models from OpenAI and Anthropic.
Even more stunning, Meta's leadership has apparently discussed licensing Gemini from Google on a temporary basis until Avocado is up to snuff, according to the NYT report. Yes, that's Google, Meta's arch-rival in the AI race (not to mention their competition in advertising, online video, and smart glasses).
These two companies are not frenemies engaging in a bit of co-opetition. And the notion of Mark Zuckerberg asking Google's Sundar Pichai for permission to use Gemini is almost impossible to imagine. (The NYT report, which is based on anonymous sources, isn't clear whether Meta executives discussed the plan internally or discussed it with Google, but either way, it's a damning indictment of Meta's AI efforts.)
Then again, Google has already seen another of its biggest rivals humbled on the AI battlefield, when Apple announced it would use Gemini for Siri.—
AO