OpenAI's highly-anticipated o3 "reasoning" AI model, unveiled in December, has had its estimated computing costs revised by the Arc Prize Foundation, the organization behind the ARC-AGI benchmark. The new estimates suggest that the best-performing configuration of o3, o3 high, could cost a staggering $30,000 per task, significantly higher than the initial estimate of $3,000.

The revision is notable not only for its magnitude but also for the insight it provides into the potential costs associated with deploying highly capable AI models. OpenAI has yet to officially price o3, but the Arc Prize Foundation believes that the pricing of OpenAI's o1-pro model, its most expensive to date, could be a reasonable proxy.

The Arc Prize Foundation's revised estimate is based on the amount of computing resources required to run o3 high. According to the organization, o3 high uses 172 times more computing resources than o3 low, the lowest-computing configuration of o3, to tackle ARC-AGI. This level of resource intensity could translate to a hefty price tag, especially for enterprise customers.

Rumors have been circulating about OpenAI's plans to introduce pricey plans for enterprise customers. In early March, The Information reported that the company may be planning to charge up to $20,000 per month for specialized AI "agents," such as a software developer agent. While these prices may seem steep, some argue that they are still lower than what a typical human contractor or staffer would command.

However, AI researcher Toby Ord points out that the efficiency of these models is also a crucial factor to consider. For instance, o3 high required 1,024 attempts at each task in ARC-AGI to achieve its best score, which could impact its overall cost-effectiveness. This raises important questions about the long-term viability of these models and their potential impact on businesses and industries.

The revised estimate also highlights the challenges associated with scaling and deploying highly capable AI models. As the technology continues to advance, it's essential to consider the economic and environmental implications of these models, including their energy consumption and carbon footprint.

In conclusion, the revised estimate of o3's computing costs serves as a reminder of the complexities and uncertainties surrounding the development and deployment of highly capable AI models. As the AI landscape continues to evolve, it's crucial to stay informed about the latest developments and their potential implications for businesses, industries, and society as a whole.