Meta’s Massive AI Spend Yields Underwhelming Results? Next-Gen Model Delayed as Performance Lags Peers

As U.S. technology giants aggressively deploy capital to expand their artificial intelligence visions, Mark Zuckerberg’s Meta appears to be trailing behind key rivals such as Google, OpenAI, and Anthropic, despite a massive surge in spending.

Meta Delays Launch of "Avocado" Model

In December, reports surfaced that Meta was developing a new text-based AI model codenamed "Avocado," with a planned release for the first quarter. However, according to a *New York Times* report on Thursday citing sources familiar with the matter, Meta has pushed back the code release for "Avocado" from March to at least May.

The delay suggests that Meta's record-breaking investments have yet to yield the desired competitive edge. In January, Meta established a staggering capital expenditure plan of $115 billion to $135 billion for this year—nearly double the $72 billion spent last year—to pursue "Superintelligence," the stage where AI surpasses human intelligence. Only months later, the company has been forced to postpone its flagship model.

Performance Struggles Against Industry Leaders

Internal testing has reportedly been the primary driver for the delay. Sources indicate that while the "Avocado" model was developed over several months, its performance in reasoning, coding, and writing tasks lags behind the leading models from Google, OpenAI, and Anthropic.

Specifically, while "Avocado" outperforms Meta's previous iterations and Google’s Gemini 2.5 (released last March), it reportedly fails to match the capabilities of Gemini 3.0, which Google debuted in November.

Ambition Remains Despite Setbacks

Despite the delay, Meta remains steadfast in its strategic direction. Responding to inquiries, Meta spokesperson Dave Arnold stated: "Our next product will be excellent, but more importantly, it will demonstrate the rapid momentum we are gaining. Throughout the coming year, we will continuously launch new models and pioneer innovative frontiers." He added that the company looks forward to unveiling the results of its R&D "very soon."

In a move highlighting the current pressure on the division, industry insiders report that Meta’s AI leadership has discussed the possibility of temporarily leasing Google’s Gemini system to power its AI products, though a final decision has not yet been reached.