In a recent post on X (formerly Twitter), OpenAI’s CEO Sam Altman shared an intriguing update on the progress of the company’s latest AI model.
“I heard o2 gets 105% on GPQA,”
Altman wrote, cryptically hinting at what could be a groundbreaking advancement in artificial intelligence performance.
i heard o2 gets 105% on GPQA
— Sam Altman (@sama) November 2, 2024
While Altman didn’t provide further details, the phrase “o2” is widely believed to refer to a new iteration or variant of OpenAI’s o1 model. Meanwhile, “GPQA” is speculated to stand for “General Purpose Question Answering,” a benchmark test used to evaluate the effectiveness of AI models in answering a wide array of questions accurately and comprehensively.
If Altman’s message is accurate, a score of 105% would signify that the model not only met but exceeded prior standards of performance on the GPQA benchmark. Typically, scores above 100% suggest that a model is surpassing established expectations—possibly excelling in challenging or nuanced question categories previously difficult for AI to navigate.
Such an accomplishment would be a significant milestone in the development of artificial intelligence, pushing AI systems closer to human-like reasoning and understanding. It also underscores OpenAI’s commitment to continuous improvement, fine-tuning models to deliver high accuracy and reliability across diverse question-answering contexts. Altman’s statement fuels ongoing speculation that a series of small yet powerful advancements, rather than one giant leap, will lead to Artificial General Intelligence (AGI).
As the race for AGI continues, OpenAI’s latest benchmark success reinforces its position at the forefront of AI innovation, indicating that we may see increasingly sophisticated AI tools emerge sooner than anticipated.
