Meta's AI President Alexandr Wang stated in an announcement to employees, based on information obtained from within the company, that they are developing a new AI model codenamed 'Watermelon', which he claims has reached a level equal to OpenAI's flagship model GPT-5.5. Wang noted at an internal company meeting that this achievement was hit on certain benchmarks that have become the industry standard for comparing AI models, emphasizing that it significantly strengthens the company's position in the AI race. It was stated that this new model, following the company's previous model known internally by the codename 'Avocado' called Muse Spark, increased the computing power used during the training phase multiple times compared to the previous generation, thus providing a performance increase.
Company owner Mark Zuckerberg has been pursuing a highly aggressive investment strategy recently to close the gap with giants like OpenAI, Google, and Anthropic in the field of artificial intelligence, and maintains a persistent stance on this issue. Zuckerberg, who restructured the AI division as 'Superintelligence Labs' last year, assigned Wang to this critical role to get ahead in this race and asked him to take the company's AI capabilities to the next level. While it is noted that the results achieved by Wang so far could be the clearest evidence that Zuckerberg's intense resource and human capital investments are starting to bear fruit, it is predicted that maintaining the gained position will be difficult given the pace of the race.
Tech giant Meta has not refrained from allocating astronomical resources to hardware, data centers, and talented engineering staff to reach this goal; infrastructure expenditures are expected to be realized between 125 billion and 145 billion dollars for this year alone. The company has launched a serious mobilization to create the most brainy team in the sector by making irresistible offers on salaries and packages to recruit the world's best AI researchers. The TBD group, an elite research team under Wang's management, and recent hardware-focused projects reveal the company's ambition to be a stakeholder not only in software but also in AI hardware.
On the other hand, OpenAI, the biggest competitor in the race, released the powerful chatbot model GPT-5.5 in April of this year, followed by the introduction of GPT 5.6 last month, described as the most powerful model the company has developed to date. However, the fact that the GPT 5.6 model has not yet been released for general use upon the request of the US government allows Meta to gain a relative advantage in the race with its 'Watermelon' model, or at least seize psychological supremacy. While no official comment was made by OpenAI regarding the issue, Meta preferred to maintain its silence against requests for verification.
Meta achieved positive results in performance tests with the Muse Spark (Avocado) model it published in April, but at that time had to admit that it could not fully reach the level of its competitors OpenAI and Anthropic. If Wang's claims are confirmed by technical reviews, it will be considered a historic turning point in showing that Meta has moved from being just a follower company to taking its place among the sector leaders. In a period when AI models have started to play such a critical role, the news that Meta has developed a model that can cope with OpenAI's strongest form is considered a development that could seriously shake the balances in the sector.
이 기사에 대해 질문
답변은 이 기사만을 바탕으로 AI가 생성합니다.