Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

OpenAI's o3 model might be costlier to run than originally estimated | TechCrunch

Apr 02, 2025 - techcrunch.com
OpenAI's o3 "reasoning" AI model, initially showcased in December, has undergone a cost reassessment by the Arc Prize Foundation, which administers the ARC-AGI benchmark. The revised estimates suggest that the computing costs for the best-performing configuration, o3 high, could be as much as $30,000 per task, a significant increase from the original $3,000 estimate. This highlights the potential expense of deploying advanced AI models for specific tasks. Although OpenAI has not yet released or priced o3, the Arc Prize Foundation considers the pricing of OpenAI's o1-pro model as a reasonable proxy due to its similar computational demands.

The high costs associated with o3 high are attributed to its substantial computing resource requirements, reportedly 172 times more than the o3 low configuration. Additionally, there are speculations about OpenAI's plans to introduce expensive enterprise AI solutions, potentially charging up to $20,000 per month for specialized AI agents. Despite these costs, some argue that AI models may still be more economical than human contractors, though their efficiency remains in question. For instance, o3 high required 1,024 attempts per task to achieve optimal performance in ARC-AGI.

Key takeaways:

  • The cost of OpenAI's o3 AI model for solving a single ARC-AGI problem has been revised from $3,000 to potentially $30,000 per task.
  • The Arc Prize Foundation suggests that OpenAI's o1-pro model pricing is a reasonable proxy for estimating o3's costs.
  • o3 high, the most computing-intensive configuration, used 172 times more computing resources than o3 low for ARC-AGI tasks.
  • There are rumors that OpenAI may charge up to $20,000 per month for specialized AI agents for enterprise customers.
View Full Article

Comments (0)

Be the first to comment!