Understanding the Situation
OpenAI’s recent AI model, o3, was initially showcased with promising capabilities through a partnership with ARC-AGI. However, new evaluations have revealed that the costs associated with using o3 are much higher than originally thought. This raises important questions about the affordability and efficiency of advanced AI models.
Key Details
- The Arc Prize Foundation revised the computing cost for o3 high from about $3,000 to approximately $30,000 per task.
- OpenAI has not yet set a price for o3, but comparisons are being made to its o1-pro model, the most expensive one available.
- The computing power required for o3 high is significantly greater, using 172 times more resources than the lowest configuration, o3 low.
- There are rumors that OpenAI may introduce high pricing plans for enterprise customers, potentially charging up to $20,000 per month for specialized AI agents.
The Bigger Picture
These developments highlight the potential financial challenges that could come with deploying advanced AI technologies. While some might argue that these costs are still lower than hiring human contractors, the efficiency of AI remains in question. For instance, o3 high required 1,024 attempts for optimal performance on tasks. As AI continues to evolve, understanding its cost-effectiveness and efficiency will be crucial for businesses considering its integration.











